WBGT-based provincial resident living electricity consumption prediction method, system and device, and storage medium

Abstract

The invention belongs to the technical field of electricity consumption prediction, and particularly provides a WBGT-based provincial resident living electricity consumption prediction method, system and device and a storage medium, and the method comprises the steps: collecting the multi-source original data of a provincial scale in a target region, calculating the hourly wet bulb and black bulb temperature, and generating a WBGT feature index; building a bidirectional fixed effect panel model containing a province fixed effect and a time fixed effect by taking provincial resident living electricity consumption data as dependent variables and WBGT characteristic indexes and social economic data as independent variables; training the constructed model through a parameter estimation method to obtain a trained prediction model; and defining a climate-social economic scene to be predicted, inputting the WBGT characteristic index and social economic data in the scene into the trained prediction model, and calculating to obtain a resident living electricity consumption estimated value. According to the method, the WBGT is adopted to carry out attribution analysis or scene simulation on the historical electricity consumption, and the estimation precision and reliability are improved.

Description

Technical Field

[0001] This invention belongs to the field of electricity consumption forecasting technology, specifically, it relates to a method, system, device and storage medium for forecasting provincial residential electricity consumption based on WBGT. Background Technology

[0002] Current methods for analyzing, attributing, and simulating provincial residential electricity consumption typically rely on macroeconomic statistics and simple temperature indicators. These methods are significantly insufficient in accurately quantifying the impact of specific climatic conditions on electricity consumption, with the following specific shortcomings: 1) Existing analytical methods cannot accurately assess the true impact of extreme hot and humid weather events on electricity consumption: Accurately assessing the load pressure on the power grid caused by a single heat wave or a series of heat waves requires a tool that can isolate the comprehensive impact of meteorological conditions from the complex socio-economic context. The actual cooling needs of the human body are determined by multiple factors, including temperature, humidity, wind speed, and radiation, but traditional temperature indicators (such as average temperature and maximum temperature) cannot fully represent this comprehensive heat stress. Currently, there is a lack of a method based on human thermophysiology that can perform high-precision "attribution analysis" of historical electricity consumption. This forces energy management departments and power companies to rely on rough statistics or single temperature indicators when conducting post-disaster assessments, power grid resilience analyses, or developing demand-side response strategies. They struggle to quantify the marginal effects of different levels of hot and humid weather on electricity consumption, thus failing to provide refined data support for infrastructure planning and electricity pricing policies.

[0003] 2) Existing assessment models are biased in quantifying the contribution of climate conditions and neglect the heterogeneous moderating effect of economic development: Significant and persistent differences exist among provinces in terms of economic development levels, industrial structures, and residents' electricity consumption habits (provincial heterogeneity). Simultaneously, nationwide technological advancements and policy regulations create shared temporal trends. These factors, along with meteorological conditions, influence electricity consumption. More importantly, the inventors' preliminary analysis revealed a systematic difference in the intensity of the impact of climate factors (such as temperature) on electricity consumption across regions with different income levels (e.g., low, medium, and high GDP groups). Conventional statistical analysis models (such as simple regression or panel models that do not consider this heterogeneity) cannot effectively capture and quantify this "heterogeneity effect" moderated by development stage. Therefore, when assessing similar "differential impacts of the same heat wave event on provinces with different development levels," existing models cannot provide accurate insights, leading to a lack of targeted policy recommendations.

[0004] 3) Currently, there is a lack of tools for highly reliable simulation of electricity demand under different climate scenarios: To formulate long-term energy strategies and power infrastructure investment plans, policymakers need to know: how would electricity consumption change if future summers returned to early 2000 levels, or became more like 2010? This counterfactual scenario simulation requires establishing a stable, clearly defined core relationship. Current methods lack a rigorously empirically tested model that characterizes the stable quantitative relationship between human thermal comfort and electricity consumption behavior. Therefore, analyses such as assessing the impact of climate change on electricity consumption growth over the past two decades or simulating electricity consumption under current socioeconomic conditions based on specific historical climate scenarios lack robustness and persuasiveness, making it difficult to support major strategic decisions. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method, system, device and storage medium for predicting provincial residential electricity consumption based on WBGT. By using WBGT, a comprehensive thermal stress index, the accuracy and reliability of the prediction are improved when performing attribution analysis or scenario simulation of historical electricity consumption.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a provincial residential electricity consumption prediction method based on WBGT, comprising the following steps: Step 1: Data Collection: Collect multi-source raw data at the provincial scale within the target area, including meteorological data, provincial residential electricity consumption data, and socio-economic data. Calculate hourly wet-bulb black-bulb temperature and aggregate hourly WBGT according to a preset time scale to generate WBGT feature index. Preprocess the multi-source raw data to obtain a standardized dataset. Step 2: Construction of a two-way fixed effects panel model: Using the provincial residential electricity consumption data preprocessed in Step 1 as the dependent variable, and the WBGT characteristic indicators and preprocessed socioeconomic data as independent variables, a two-way fixed effects panel model including provincial fixed effects and time fixed effects is constructed. Step 3, Model Training: Using the standardized dataset from Step 1 as the training set, the two-way fixed effects panel model constructed in Step 2 is trained using parameter estimation methods to obtain the trained prediction model. Step 4: Estimate electricity consumption based on the trained prediction model: Define the climate-socioeconomic scenario to be predicted, input the WBGT characteristic index and socioeconomic data under the scenario into the prediction model trained in Step 3, and calculate the estimated value of residential electricity consumption.

[0007] In a preferred embodiment, the specific types of multi-source raw data in step 1 include: 1) Meteorological data: Hourly data for each province in the target area over a historical period, including dry-bulb temperature, relative humidity, wind speed and solar radiation intensity. The data source is observation data or reanalysis datasets from the China Meteorological Administration. 2) Electricity consumption and socioeconomic data: Collect panel data for the corresponding year and province, including: Dependent variable: Provincial annual residential electricity consumption; Independent variables: annual population of the province, GDP per capita, level of industrialization and urbanization rate, where the level of industrialization is the percentage of the added value of the secondary industry to GDP.

[0008] In a preferred embodiment, the formula for calculating the hourly wet-bulb black bulb temperature is as follows: ; Where Ta is the dry-bulb temperature; Tnw is the natural wet-bulb temperature; and Tg is the black-bulb temperature.

[0009] In the preferred embodiment, the estimation of the natural wet-bulb temperature is based on: Tnw=Ta-0.0026×(100-RH)×(1+0.01×Ta); Where RH represents relative humidity; The estimation of the black ball's temperature is based on: Tg = Ta + 0.0038 × SR; Where SR represents solar radiation intensity.

[0010] In a preferred embodiment, the method for obtaining the WBGT characteristic index is as follows: hourly WBGT is aggregated according to summer, which includes the three months from June to August, to generate the summer average WBGT for each province and each year.

[0011] In the preferred embodiment, step 1, the preprocessing of multi-source raw data includes natural logarithmic transformation of provincial residential electricity consumption data, population data, and per capita GDP data.

[0012] In the preferred embodiment, the mathematical expression of the two-way fixed-effects panel model in step 2 is: ; in: i Indicates the first i Provinces, t Indicates the first t Year; For the first i Province t The natural logarithm of annual residential electricity consumption is the dependent variable in the model. For the first i Province t WBGT characteristic index of the yearβ 1 represents the coefficient to be estimated. , The first i Province t The natural logarithm of annual population and GDP per capita; , These are precipitation and humidity, respectively. α For the intercept term; For provincial fixed effects; This is a time-fixed effect; For random disturbance terms; β 1-β7 represents the corresponding coefficient to be estimated.

[0013] In the preferred embodiment, in step 2, an interaction term between WBGT and GDP per capita is introduced into the two-way fixed effects panel model to construct an interaction term model. Through interaction term or group analysis, the economic development heterogeneity of the impact of WBGT on residential electricity consumption is quantified for prediction of electricity consumption within the province. The expression of the interaction term model is as follows: ; in: i Indicates the first i Provinces, t Indicates the first t Year; For the first i Province t The natural logarithm of annual residential electricity consumption is the dependent variable in the model. For the first i Province t WBGT characteristic index of the year β 1 represents the coefficient to be estimated. , The first i Province t The natural logarithm of annual population and GDP per capita; , This is an interaction term used to examine the nonlinear regulatory effect of economic development on the impact of WBGT on electricity consumption; , These are precipitation and humidity, respectively. α For the intercept term; For provincial fixed effects; This is a time-fixed effect; For random disturbance terms; β 1 ~β10 For the corresponding coefficients to be estimated.

[0014] In the preferred embodiment, step 3, the specific process of model training includes: 1) Training set partitioning: Divide the standardized dataset from step 1 into a training set and a validation set; 2) Parameter estimation: Using the least squares dummy variable method or internal transformation method, the coefficients α, β1~β7, and... of the two-way fixed effects panel model were estimated. , Make an estimate.

[0015] In the preferred embodiment, in step 3, during the training phase, when training the interaction term model, provinces are divided into high economic development level group, medium economic development level group, and low economic development level group according to the median GDP per capita. The model is trained separately for each subsample according to the interaction term model to obtain the estimated coefficients corresponding to the WBGT feature indicators of each group, thereby quantifying the heterogeneity differences.

[0016] In the preferred embodiment, step 3 further includes model validation, which involves inputting the divided validation set into the trained prediction model, calculating the average absolute error or average absolute percentage error between the predicted value and the actual residential electricity consumption, and comparing it with a set threshold. If the average absolute error or average absolute percentage error is less than the set threshold, the model is deemed to have passed validation; otherwise, the model parameters are readjusted and training is repeated.

[0017] In the preferred embodiment, in step 4, the climate-socioeconomic scenario to be predicted is defined as including climate scenario and socioeconomic scenario, and the WBGT characteristic indicators and socioeconomic data corresponding to the scenario are substituted into the trained prediction model for prediction.

[0018] In the preferred embodiment, in step 4, the output value of the trained prediction model is... ln ( Electricity consumption The electricity consumption is calculated through exponential operations.

[0019] This invention also provides a residential electricity consumption forecasting system based on WBGT, used to execute the aforementioned provincial residential electricity consumption forecasting method based on WBGT, comprising: Data preparation and processing module: Collects multi-source raw data at the provincial scale within the target area, calculates hourly wet-bulb black-bulb temperature, aggregates hourly WBGT according to a preset time scale, generates WBGT feature index, preprocesses the multi-source raw data, and obtains a standardized dataset; Model building and training module: Construct a two-way fixed effects panel model and train it to obtain a trained prediction model; Prediction and Application Module: Predicts electricity consumption based on the trained prediction model; Results output module: Outputs the predicted results in a visual format.

[0020] The present invention also provides an electronic device, characterized in that it includes a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the above-described method for predicting provincial residential electricity consumption based on WBGT.

[0021] The present invention also provides a computer-readable storage medium storing computer instructions for causing the computer to execute the above-described method for predicting provincial residential electricity consumption based on WBGT.

[0022] The present invention provides a provincial-level residential electricity consumption prediction method, system, device, and storage medium based on WBGT, which has the following beneficial effects: 1. For the first time, in the prediction / attribution analysis model of residential electricity consumption at the provincial and above scales, wet-bulb black-sphere temperature (WBGT) or its derivative indicators (such as summer average WBGT and hours exceeding WBGT) are used as key explanatory variables characterizing climate conditions, replacing traditional indicators that only consider temperature (such as average temperature and degree-days). This utilizes the scientific principle that WBGT can more comprehensively reflect the actual thermal sensation of the human body, fundamentally solving the prediction bias problem caused by insufficient physiological mechanisms in existing technologies. This is the theoretical basis and primary technical feature of this invention for achieving high-precision prediction.

[0023] This invention employs the WBGT (Weighted Booster Temperature) comprehensive thermal stress index, whose physical content is highly consistent with the actual thermal comfort sensation of the human body and the resulting cooling behavior. Compared to existing technologies that only consider dry-bulb temperature, this invention fundamentally overcomes the deficiency of systematically underestimating electricity demand in high-temperature and high-humidity environments. For example, when facing "sauna-like" weather, existing technology models may be insufficiently responsive, while this invention, by incorporating key factors such as humidity, can more sensitively and accurately reflect the resulting surge in electricity load. Therefore, when conducting attribution analysis of historical electricity consumption or performing scenario simulations, the prediction accuracy and reliability of this invention are far superior to existing technologies.

[0024] 2. This invention addresses the problem of existing technologies using an "average effect" to mask "real differences" by identifying and quantifying the heterogeneity of the WBGT effect among GDP groups. This enables the invention to clearly indicate that the impact of nationwide heat waves is far greater on high-income provinces than on low-income provinces when assessing their impact, thus achieving a leap from "extensive assessment" to "precise diagnosis."

[0025] 3. This invention constructs a more rigorous and in-depth econometric model by introducing provincial and time-fixed effects and further examining heterogeneity. This design not only effectively eliminates confounding factors but also reveals the underlying principle that "economic development level is a key variable regulating the climate-energy relationship." Compared with existing technologies, this invention outputs not a general coefficient but a set of more insightful conclusions that reflect complex socio-economic-climate interactions, greatly enhancing its decision support value.

[0026] 3. Simulating differentiated electricity consumption responses based on socioeconomic conditions at different development levels under the same historical climate scenario. Compared to existing technologies, this invention significantly enhances the credibility of results when simulating counterfactual scenarios (e.g., assessing the impact of past climate change or simulating energy demand under different future climate pathways). This enables power companies and energy planning departments to develop more forward-looking and scientific infrastructure investment plans and climate adaptation strategies based on the methods of this invention, with decision support value far exceeding that of existing technologies. Attached Figure Description

[0027] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart of the process of this invention; Figure 2 It is a forest plot comparing the estimated coefficients of each variable grouped by income; Figure 3 It is a descriptive statistic of a two-way fixed-effects panel regression model; Figure 4 It is a descriptive statistic of a two-way fixed-effects panel regression model that incorporates a multinomial interaction term. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0029] Example 1: A provincial-level residential electricity consumption forecasting method based on WBGT mainly includes three core stages: Data preparation and WBGT calculation phase: Acquire raw data from multiple sources and calculate the core meteorological indicator, wet-bulb black-bulb temperature (WBGT).

[0030] Model building and training phase: Using historical data, a fixed-effects panel model is built and trained to establish a stable quantitative relationship between WBGT and electricity consumption.

[0031] Prediction and Application Phase: New, user-specified scenario data is fed into the trained model to obtain accurate electricity consumption prediction results.

[0032] like Figure 1 As shown, it includes the following steps: Step 1: Data Collection: The data preparation and processing module collects multi-source raw data at the provincial scale within the target area, including meteorological data, provincial residential electricity consumption data, and socio-economic data. The hourly wet-bulb black-bulb temperature is calculated, and the hourly WBGT is aggregated according to a preset time scale to generate WBGT feature indicators. The multi-source raw data is preprocessed to obtain a standardized dataset.

[0033] 1. Data Collection: 1) Meteorological Data: Obtain hourly meteorological observations or reanalysis data for the target region (e.g., provinces in China) over a historical period (e.g., 2000-2024). Required meteorological elements include: dry-bulb temperature (°C), relative humidity (%), wind speed (m / s), and solar radiation intensity (W / m²). Data sources can include the China Meteorological Administration observation dataset, reanalysis datasets such as ERA5, etc.

[0034] 2) Electricity consumption and socioeconomic data: Collect panel data for the corresponding year and province, including: Dependent variable: Provincial annual residential electricity consumption (100 million kWh).

[0035] Independent variables: annual population of the province (ten thousand people), GDP per capita (yuan / person), level of industrialization (usually expressed as the proportion of secondary industry added value to GDP, %), and urbanization rate (%).

[0036] 2. WBGT Index Calculation and Aggregation Calculate the hourly wet-bulb black-bulb temperature and aggregate the hourly WBGT according to a preset time scale to generate WBGT characteristic indices. The formula for calculating the hourly wet-bulb black-bulb temperature is as follows: ; Where Ta is the dry-bulb temperature; Tnw is the natural wet-bulb temperature; and Tg is the black-bulb temperature.

[0037] Tnw and Tg can be estimated using empirical formulas based on temperature, humidity, wind speed, and radiation.

[0038] In this embodiment, the estimation of the natural wet-bulb temperature is based on: Tnw=Ta-0.0026×(100-RH)×(1+0.01×Ta); Here, RH represents relative humidity.

[0039] The estimation of the black ball's temperature is based on: Tg = Ta + 0.0038 × SR; Where SR represents solar radiation intensity.

[0040] The calculated hourly WBGT data are aggregated by summer (June, July, and August) to generate the summer average WBGT for each province and each year. In addition, other meaningful indicators can be calculated, such as the cumulative number of hours when summer WBGT exceeds a certain health risk threshold (e.g., 28°C or 30°C), as alternative or supplementary explanatory variables.

[0041] 3. Data preprocessing: Preprocessing of multi-source raw data includes natural logarithmic transformation of provincial residential electricity consumption data, population data, and GDP per capita data. This allows the model coefficients to be interpreted as elasticity (percentage change). For ratio variables (such as industrialization level), the original value can be retained or a logical transformation can be performed.

[0042] Check and handle missing and outlier values ​​to ensure data quality.

[0043] Step 2: Construction of a Two-Way Fixed Effects Panel Model: Using the preprocessed provincial residential electricity consumption data from Step 1 as the dependent variable, and the WBGT characteristic indicators and preprocessed socioeconomic data as independent variables, a two-way fixed effects panel model incorporating provincial and time fixed effects is constructed. The model is built and trained using the model building and training module.

[0044] The mathematical expression for the two-way fixed-effects panel model is: ; in: i Indicates the first i Provinces, t Indicates the first t Year; For the first i Province t The natural logarithm of annual residential electricity consumption is the dependent variable in the model. For the first i Province t WBGT characteristic index of the year β 1 represents the coefficient to be estimated. , The first i Province t The natural logarithm of annual population and GDP per capita; , These are precipitation and humidity, respectively. α For the intercept term; Province fixed effect is used to capture the characteristics of each province that do not change over time (such as geographical location, long-established electricity consumption habits, etc.). This is a time-fixed effect used to capture time trends that are common to all provinces (such as nationwide technological progress, macroeconomic policies, and international energy price shocks). For random disturbance terms; β 1-β7 represents the corresponding coefficient to be estimated.

[0045] Among them, β1 is the coefficient of most interest in this invention, which reflects the percentage change in residential electricity consumption for every 1°C increase in the average WBGT in summer after controlling for other factors.

[0046] An interaction term between WBGT and GDP per capita is introduced into a two-way fixed-effects panel model to construct an interaction term model. Through interaction term or group analysis, the economic development heterogeneity of the impact of WBGT on residential electricity consumption is quantified for prediction of electricity consumption within the province. The expression of the interaction term model is as follows: ; in: i Indicates the first i Provinces, t Indicates the first t Year; For the first i Province t The natural logarithm of annual residential electricity consumption is the dependent variable in the model. For the first i Province t WBGT characteristic index of the year β 1 represents the coefficient to be estimated. , The first i Province t The natural logarithm of annual population and GDP per capita; , This is an interaction term used to examine the nonlinear regulatory effect of economic development on the impact of WBGT on electricity consumption; , These are precipitation and humidity, respectively. α For the intercept term; For provincial fixed effects; This is a time-fixed effect; For random disturbance terms; β 1 ~β10 For the corresponding coefficients to be estimated.

[0047] Step 3, Model Training: Using the standardized dataset from Step 1 as the training set, the two-way fixed effects panel model constructed in Step 2 is trained using parameter estimation methods to obtain the trained prediction model.

[0048] The specific process of model training includes: 1) Training set partitioning: Divide the standardized dataset from step 1 into a training set and a validation set; 2) Parameter estimation: Using the least squares dummy variable method or internal transformation method, the coefficients α, β1~β7, and... of the two-way fixed effects panel model were estimated. , Make an estimate.

[0049] Use historical data (e.g., data from 2000-2020) as the training set.

[0050] Model parameters are estimated using the least squares dummy variable method or internal transformation method in econometric software (such as Stata, R, and Python's linearmodels library). These methods are standard techniques for estimating fixed effects models.

[0051] During the training process, provinces are divided into high economic development level group, medium economic development level group, and low economic development level group according to the median GDP per capita. The model is trained separately for each subsample to obtain the estimated coefficients of the WBGT feature indicators corresponding to each group, thus quantifying the heterogeneity differences.

[0052] After training, a set of definite coefficient estimates (β1) is obtained. β2 ...β10 (and fixed effects values ​​for provinces and years.)

[0053] To verify and quantify the heterogeneity of the WBGT effect at different stages of economic development, this invention employs a grouped modeling approach. Specifically, based on the historical median per capita GDP of each province, the sample is divided into subsamples of "high-income," "middle-income," and "low-income," and model estimations are performed for each subsample separately. Figure 2 As shown.

[0054] Figure 2This paper presents the estimated coefficients (points) and confidence intervals (horizontal lines) of WBGT and other control variables on per capita residential electricity consumption in low, middle, and high economic development groups. A key finding is that the coefficient of WBGT (the horizontal axis corresponding to the coefficients of 'WBGT' or 'Leg GDP pc' in the graph) is significantly positive and has the largest value in the high-income group, while it is smaller in the low- and middle-income groups. This visually and conclusively demonstrates that the strength of the impact of WBGT on electricity consumption is affected by the level of economic development, validating the necessity and correctness of the heterogeneity analysis in this invention.

[0055] By comparing the coefficients of WBGT in different subsamples, it is possible to quantitatively reveal how the impact of WBGT on electricity consumption changes with the stage of economic development. For example... Figure 2 As shown, the empirical results of the model of this invention clearly show that the WBGT coefficient (β1) of the high-income group is significantly greater than that of the middle- and low-income group, proving the effectiveness and necessity of this method.

[0056] Input the divided validation set (such as data from 2021 to 2024) into the trained prediction model, calculate the mean absolute error or mean absolute percentage error between the predicted value and the actual residential electricity consumption, and compare it with the set threshold. If the mean absolute error or mean absolute percentage error is less than the set threshold, the model is considered to have passed the validation; otherwise, the model parameters are readjusted and the training is repeated.

[0057] To demonstrate the fitting effect of the constructed two-way fixed-effects panel model and interaction term model of this invention, the following verifications were performed: The model was estimated using the reghdfe command, with per capita electricity consumption as the dependent variable and wbgt as the core independent variable. It controlled for per capita GDP (In_gdppc), urbanization rate (urban_rate), and other environmental variables (pre, rhu, win, sir). The model incorporated fixed effects for province (provinceld) and year (year) to control for unobservable province specificity and time trends, and used province-level cluster robust standard error (vce(clusterprovinceld)) to handle heteroscedasticity and within-group correlation, which improved the reliability of the estimates.

[0058] like Figure 3As shown, the model converged after two iterations, indicating that the numerical calculation was stable. Regarding the significance of the coefficients of the key independent variables, the coefficient of wbgt was 38.05133 across all samples, significant at the 1% significance level (p = 0.002). The wbgt coefficients for the high GDP group, medium GDP group, and low GDP group were 47.17 (significant at the 1% significance level), 19.28 (significant at the 10% significance level), and 29.16 (significant at the 5% significance level), respectively, indicating that this variable has a significant positive impact on per capita electricity consumption in all groups and can be used to predict the dependent variable, per capita electricity consumption.

[0059] The model incorporates fixed effects from 30 provinces and 23 years. The province fixed effects are completely absorbed due to nesting within clusters, resulting in zero degrees of freedom. The year fixed effects retain 22 coefficients, effectively controlling for unobservable heterogeneity. In summary, the model performs well in controlling for spatiotemporal fixed effects and cluster robustness standard errors, and the significant effect of the core variable wbgt supports the research hypotheses.

[0060] like Figure 4 As shown, by introducing the interaction term and nonlinear term between wbgt_jjas and ln_gdppc, a core mechanism obscured by a simple linear relationship is revealed: the impact of climate variables is not static, but is significantly moderated by the level of economic development. The main effect of wbgt is significant (p=0.039); the quadratic interaction term between wbgt and GDP (c.wbgt_jjas#c.ln_gdppc#c.ln_gdppc) is significant (p=0.024); their primary interaction term (c.wbgt_jjas#c.ln_gdppc) is also significant (p=0.038). These significant terms collectively indicate that the impact of the climate variable (wbgt) on the dependent variable (per capita residential electricity consumption) is not fixed, but depends on the local level of economic development (GDP), and this dependence is nonlinear. This means that the direction and intensity of the impact of global warming may be completely different in low-income and high-income regions. Furthermore, the overall goodness of fit of the model (R-squared=0.9714) and the significant within-group explanatory power (WithinR-sq.=0.4436) indicate that the model can capture the complex structure in the data very well.

[0061] Step 4: Estimate electricity consumption based on the trained prediction model: Define the climate-socioeconomic scenario to be predicted, including the climate scenario and the socioeconomic scenario. The prediction and application module inputs the WBGT characteristic index and socioeconomic data under the scenario into the prediction model trained in Step 3 to calculate the estimated value of residential electricity consumption.

[0062] The user sets a scenario that needs to be predicted. For example: Differential attribution analysis: quantifying the differentiated electricity consumption impacts of the same nationwide heat wave (such as the summer of 2022) on high-, middle-, and low-income provinces.

[0063] Differentiated Scenario Simulation: How will electricity consumption differentiate if, under the socioeconomic conditions of high-, middle-, and low-income provinces, they encounter the WBGT situation of 2013 (a historically exceptionally hot summer)? Substitute the WBGT value and socioeconomic variable values ​​under the target scenario into the trained model.

[0064] Key point: Province fixed effects in the model and year fixed effect It needs to be handled correctly according to the scenario. For example, when simulating the situation in various provinces in 2023, The estimated effect for 2023 (if after the training period) or a reasonable extrapolation of the effect for the most recent year should be taken. Then take the fixed effect value for the corresponding province.

[0065] By multiplying the input data by the model coefficients and incorporating fixed effects, the estimated value of ln (electricity consumption) is calculated. The result is then converted back to the original unit of electricity consumption (billion kilowatt-hours) through exponential operations.

[0066] Results output module: Outputs the predicted results in a visual format and can trigger the following applications: Grid Resilience Report: Generates a quantitative assessment report on the impact of historical extreme hot and humid weather events on grid load.

[0067] Policy simulation platform: As a backend engine, it is embedded in the decision support system, allowing users to input different climate and socioeconomic assumptions and view the electricity consumption simulation results in real time.

[0068] Risk warning: When the simulation results show that the electricity consumption surges and exceeds a certain threshold, the system can issue a warning to the relevant departments.

[0069] Through the above four steps, this invention achieves a complete technical closed loop for residential electricity consumption, from data preparation and model building to accurate prediction and application, effectively solving many defects of the existing solutions described in the background art.

[0070] Example 2: This embodiment provides a residential electricity consumption forecasting system based on WBGT, used to execute a provincial residential electricity consumption forecasting method based on WBGT as described in Embodiment 1, including: Data preparation and processing module: Collects multi-source raw data at the provincial scale within the target area, calculates hourly wet-bulb black-bulb temperature, aggregates hourly WBGT according to a preset time scale, generates WBGT feature index, preprocesses the multi-source raw data, and obtains a standardized dataset; Model building and training module: Construct a two-way fixed effects panel model and train it to obtain a trained prediction model; Prediction and Application Module: Predicts electricity consumption based on the trained prediction model; Results output module: Outputs the predicted results in a visual format.

[0071] Example 3: This embodiment provides an electronic device, including a memory and a processor, which are communicatively connected. The memory stores computer instructions, and the processor executes the computer instructions to perform the provincial residential electricity consumption prediction method based on WBGT described in Embodiment 1.

[0072] Example 4: This embodiment provides a computer-readable storage medium storing computer instructions for causing the computer to execute the provincial residential electricity consumption prediction method based on WBGT described in Embodiment 1.

[0073] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A provincial-level residential electricity consumption forecasting method based on WBGT, characterized in that, Includes the following steps: Step 1: Data Collection: Collect multi-source raw data at the provincial scale within the target area, including meteorological data, provincial residential electricity consumption data, and socio-economic data. Calculate hourly wet-bulb black-bulb temperature and aggregate hourly WBGT according to a preset time scale to generate WBGT feature index. Preprocess the multi-source raw data to obtain a standardized dataset. Step 2: Construction of a two-way fixed effects panel model: Using the provincial residential electricity consumption data preprocessed in Step 1 as the dependent variable, and the WBGT characteristic indicators and preprocessed socioeconomic data as independent variables, a two-way fixed effects panel model including provincial fixed effects and time fixed effects is constructed. Step 3, Model Training: Using the standardized dataset from Step 1 as the training set, the two-way fixed effects panel model constructed in Step 2 is trained using parameter estimation methods to obtain the trained prediction model. Step 4: Estimate electricity consumption based on the trained prediction model: Define the climate-socioeconomic scenario to be predicted, input the WBGT characteristic index and socioeconomic data under the scenario into the prediction model trained in Step 3, and calculate the estimated value of residential electricity consumption.

2. The provincial residential electricity consumption forecasting method based on WBGT according to claim 1, characterized in that, The specific types of multi-source raw data in step 1 include: 1) Meteorological data: Hourly data for each province in the target area over a historical period, including dry-bulb temperature, relative humidity, wind speed and solar radiation intensity. The data source is observation data or reanalysis datasets from the China Meteorological Administration. 2) Electricity consumption and socioeconomic data: Collect panel data for the corresponding year and province, including: Dependent variable: Provincial annual residential electricity consumption; Independent variables: annual population of the province, GDP per capita, level of industrialization and urbanization rate, where the level of industrialization is the percentage of the added value of the secondary industry to GDP.

3. The provincial residential electricity consumption forecasting method based on WBGT according to claim 1, characterized in that, The formula for calculating the hourly wet-bulb black bulb temperature is as follows: ； Where Ta is the dry-bulb temperature; Tnw is the natural wet-bulb temperature; and Tg is the black-bulb temperature.

4. The provincial residential electricity consumption forecasting method based on WBGT according to claim 3, characterized in that, The estimation of the natural wet-bulb temperature is based on: Tnw=Ta-0.0026×(100-RH)×(1+0.01×Ta); Where RH represents relative humidity; The estimation of the black ball's temperature is based on: Tg = Ta + 0.0038 × SR; Where SR represents solar radiation intensity.

5. The provincial residential electricity consumption forecasting method based on WBGT according to claim 1, characterized in that, The method for obtaining the WBGT characteristic index is as follows: hourly WBGT is aggregated according to summer, which includes the three months from June to August, to generate the summer average WBGT for each province and each year.

6. The provincial residential electricity consumption forecasting method based on WBGT according to claim 1, characterized in that, In step 1, the preprocessing of multi-source raw data includes natural logarithmic transformation of provincial residential electricity consumption data, population data, and per capita GDP data.

7. The provincial residential electricity consumption forecasting method based on WBGT according to claim 1, characterized in that, In step 2, the mathematical expression of the two-way fixed-effects panel model is: ； in: i Indicates the first i Provinces, t Indicates the first t Year; For the first i Province t The natural logarithm of annual residential electricity consumption is the dependent variable in the model. For the first i Province t WBGT characteristic index of the year β 1 represents the coefficient to be estimated. , The first i Province t The natural logarithm of annual population and GDP per capita; , These are precipitation and humidity, respectively. α For the intercept term; For provincial fixed effects; This is a time-fixed effect; For random disturbance terms; β 1-β7 represents the corresponding coefficient to be estimated.

8. The provincial residential electricity consumption prediction method based on WBGT according to claim 7, characterized in that, In step 2, an interaction term between WBGT and GDP per capita is introduced into the two-way fixed effects panel model to construct an interaction term model. Through interaction term or group analysis, the economic development heterogeneity of the impact of WBGT on residential electricity consumption is quantified for prediction of electricity consumption within the province. The expression of the interaction term model is as follows: ； in: i Indicates the first i Provinces, t Indicates the first t Year; For the first i Province t The natural logarithm of annual residential electricity consumption is the dependent variable in the model. For the first i Province t WBGT characteristic index of the year β 1 represents the coefficient to be estimated. , The first i Province t The natural logarithm of annual population and GDP per capita; , This is an interaction term used to examine the nonlinear regulatory effect of economic development on the impact of WBGT on electricity consumption; , These are precipitation and humidity, respectively. α For the intercept term; For provincial fixed effects; This is a time-fixed effect; For random disturbance terms; β 1 ~β10 For the corresponding coefficients to be estimated.

9. The provincial residential electricity consumption forecasting method based on WBGT according to claim 1, characterized in that, In step 3, the specific process of model training includes: 1) Training set partitioning: Divide the standardized dataset from step 1 into a training set and a validation set; 2) Parameter estimation: Using the least squares dummy variable method or internal transformation method, the coefficients α, β1~β7, and... of the two-way fixed effects panel model were estimated. , Make an estimate.

10. A provincial-level residential electricity consumption forecasting method based on WBGT according to claim 8, characterized in that, In step 3, during the training phase, when training the interaction term model, provinces are divided into high economic development level group, medium economic development level group, and low economic development level group according to the median GDP per capita. The model is trained separately for each subsample according to the interaction term model to obtain the estimated coefficients corresponding to the WBGT feature indicators of each group, thus quantifying the heterogeneity differences.

11. The provincial residential electricity consumption prediction method based on WBGT according to claim 8, characterized in that, Step 3 also includes model validation, which involves inputting the divided validation set into the trained prediction model, calculating the average absolute error or average absolute percentage error between the predicted value and the actual residential electricity consumption, and comparing it with a set threshold. If the average absolute error or average absolute percentage error is less than the set threshold, the model is deemed to have passed validation; otherwise, the model parameters are readjusted and training is repeated.

12. The provincial residential electricity consumption prediction method based on WBGT according to claim 1, characterized in that, In step 4, the climate-socioeconomic scenario to be predicted is defined as including climate scenario and socioeconomic scenario. The WBGT characteristic index and socioeconomic data corresponding to the scenario are substituted into the trained prediction model for prediction.

13. The provincial residential electricity consumption forecasting method based on WBGT according to claim 1, characterized in that, In step 4, the output value of the trained prediction model is ln ( Electricity consumption The electricity consumption is calculated through exponential operations.

14. A residential electricity consumption forecasting system based on WBGT, characterized in that, A method for predicting provincial residential electricity consumption based on WBGT as described in any one of claims 1 to 13 includes: Data preparation and processing module: Collects multi-source raw data at the provincial scale within the target area, calculates hourly wet-bulb black-bulb temperature, aggregates hourly WBGT according to a preset time scale, generates WBGT feature index, preprocesses the multi-source raw data, and obtains a standardized dataset; Model building and training module: Construct a two-way fixed effects panel model and train it to obtain a trained prediction model; Prediction and Application Module: Predicts electricity consumption based on the trained prediction model; Results output module: Outputs the predicted results in a visual format.

15. An electronic device, characterized in that, The device includes a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform a provincial residential electricity consumption prediction method based on WBGT as described in any one of claims 1 to 13.

16. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute a WBGT-based provincial residential electricity consumption prediction method as described in any one of claims 1 to 13.

Images