Electric charge income prediction method, computer device and storage medium

Abstract

The invention provides an electric charge income prediction method, a computer device and a storage medium, and the method comprises the steps: collecting model input data of multiple dimensions through a power grid enterprise marketing system, the model input data comprising a transaction mode, a power utilization type, a voltage class, and historical electricity sale quantity data and historical electricity sale unit price data corresponding to peaks and flat valleys; establishing a general prediction model corresponding to the plurality of prediction indexes; predicting the model input data by using the general prediction model corresponding to each prediction index, and obtaining the sub-model predicted electricity sale quantity and the sub-model predicted electricity sale unit price corresponding to each prediction index in the current to-be-predicted month; summarizing the sub-model predicted electricity sale quantity and the sub-model predicted electricity sale unit price to obtain the predicted electricity sale quantity and the predicted electricity sale unit price of the current to-be-predicted month; and obtaining the predicted electric charge receivable amount of the current to-be-predicted month according to the predicted electricity sale quantity and the predicted electricity sale unit price. According to the invention, the method can achieve the refined prediction of the electricity fund income, and improves the prediction precision and stability.

Description

Technical Field

[0001] The present invention relates to the technical field of electricity fee prediction, and specifically, to an electricity fee revenue prediction method, a computer device applying the electricity fee revenue prediction method, and a computer-readable storage medium applying the electricity fee revenue prediction method. Background Art

[0002] Power companies need to more accurately predict electricity revenue in order to formulate reasonable market strategies, financial plans and risk management measures. However, in reality, the accuracy of electricity revenue forecasting faces many challenges and difficulties. Existing electricity revenue forecasting technology mainly predicts monthly aggregated electricity revenue (single indicator). This method prevents users from gaining an in-depth understanding of detailed errors, limits users' grasp of more detailed indicator data, and makes it difficult to discover potential problems in operations or production, while also limiting the room for accuracy optimization.

[0003] In response to these problems, some researchers have proposed some solutions. For example, some researchers have proposed a method for predicting electricity bill revenue based on the Markov state transfer matrix. This method mainly performs monthly cash flow forecast calculations and outputs the predicted monthly cash flow data based on the visualization of historical indicator data. The predicted monthly cash flow data is input into the Markov state transfer model, and the daily bill revenue cash flow inflow is output. However, this method has some problems. First, this method only combines the Markov state transfer matrix for model prediction. The diversity of historical data cannot be used for multi-dimensional prediction. It can only filter the data in the historical data that meets the Markov state transfer matrix type and output the results. The model cannot consider the actual business data situation for multi-dimensional prediction analysis and cannot meet the actual needs and challenges of enterprises.

[0004] Therefore, a more optimized electricity bill revenue forecasting method needs to be considered. Summary of the invention

[0005] The first object of the present invention is to provide a method for predicting electricity fee income, which can realize refined prediction of electricity fee financial income and improve prediction accuracy and stability.

[0006] The second object of the present invention is to provide a computer device that can realize refined prediction of electricity fee capital income and improve prediction accuracy and stability.

[0007] The third object of the present invention is to provide a computer-readable storage medium that can achieve refined prediction of electricity fee capital income and improve prediction accuracy and stability.

[0008] In order to achieve the first object of the present invention, the electricity fee income prediction method provided by the present invention includes: collecting model input data of multiple dimensions through the marketing system of the power grid enterprise, the model input data including historical electricity sales data and historical electricity sales price data corresponding to the transaction mode, electricity consumption category, voltage level and peak and valley; establishing a general prediction model corresponding to multiple prediction indicators; using the general prediction model corresponding to each prediction indicator to predict the model input data, and obtain the sub-model predicted electricity sales and sub-model predicted electricity sales price corresponding to each prediction indicator of the current month to be predicted; summarizing the sub-model predicted electricity sales and sub-model predicted electricity sales price to obtain the predicted electricity sales and predicted electricity sales price of the current month to be predicted; obtaining the predicted electricity fee receivable amount of the current month to be predicted based on the predicted electricity sales and predicted electricity sales price.

[0009] It can be seen from the above scheme that the electricity fee income prediction method of the present invention collects model input data of multiple dimensions, uses the general prediction model corresponding to multiple prediction indicators to predict the model input data, and summarizes the predicted sub-model predicted electricity sales volume and sub-model predicted electricity unit price to obtain the predicted electricity sales volume and predicted electricity unit price of the current month to be predicted, and then obtains the electricity fee receivable amount of the current month to be predicted. It can realize multi-dimensional and multi-category analysis in combination with actual business needs. The application scope of the model calculation results is relatively wide, and more fine-grained control can be performed from the dimensions of electricity consumption category, peak and valley, and voltage level, thereby improving the accuracy and stability of the prediction.

[0010] In a further solution, after collecting model input data of multiple dimensions through the marketing system of the power grid enterprise, it also includes: distinguishing the model input data to obtain first characteristic data of a time series with trend or stability and second characteristic data of a time series without trend or stability.

[0011] It can be seen that since it is impossible to guarantee that the time series of each indicator has a good trend or stability, it is necessary to distinguish the data characteristics of different indicators in the model input data, so that more appropriate prediction methods can be adopted for indicators with different characteristics in the subsequent prediction stage.

[0012] In a further scheme, the steps of distinguishing the model input data to obtain first characteristic data of a time series with trend or stability and second characteristic data of a time series without trend or stability include: performing mean normalization processing on the model input data corresponding to each prediction indicator, and calculating the coefficient of variation; if the coefficient of variation is less than or equal to the coefficient of variation fluctuation threshold, the model input data corresponding to the prediction indicator is the first characteristic data; if the coefficient of variation is greater than the coefficient of variation fluctuation threshold, the model input data corresponding to the prediction indicator is the second characteristic data.

[0013] It can be seen that by using the coefficient of variation to distinguish between the first characteristic data of the time series with trend or stability and the second characteristic data of the time series without trend or stability, the accuracy of classification can be improved.

[0014] In a further scheme, the step of predicting the model input data using the general prediction model corresponding to each prediction indicator includes: using at least two preset time series algorithms to perform combined prediction on the first characteristic data to obtain the sub-model predicted electricity sales volume and the sub-model predicted electricity unit price corresponding to the first characteristic data; and / or performing weighted addition based on the indicator value of the same period last year and the indicator value of the previous month to obtain the sub-model predicted electricity sales volume and the sub-model predicted electricity unit price corresponding to the second characteristic data.

[0015] It can be seen that the combined prediction of the first characteristic data using at least two preset time series algorithms can make full use of the respective advantages of different algorithms and make up for their respective deficiencies, thereby achieving accurate prediction of indicator data with trend or stability. The predicted data of the second characteristic data is obtained by weighted addition of the indicator value of the same period last year and the indicator value of the previous month, which can avoid the influence of strong fluctuations of the time series on the predicted value to the greatest extent.

[0016] In a further solution, after the step of collecting model input data of multiple dimensions through the marketing system of the power grid enterprise, it also includes: sorting and cleaning the model input data.

[0017] It can be seen that since the model input data involves many complex dimensions, the original data may contain various irregularities, inaccuracies, and anomalies. Therefore, before conducting subsequent in-depth analysis, the data must be carefully organized and cleaned to ensure the quality and availability of the data.

[0018] In a further solution, the step of arranging and cleaning the model input data includes: filtering the model input data using an outlier filtering method with a preset multiple of IQR.

[0019] It can be seen that filtering the model input data through the outlier filtering method with a preset multiple IQR can avoid the impact of dirty data and abnormal data, and ensure that the data ultimately used for analysis and prediction are relatively "clean" data within a reasonable range.

[0020] In a further scheme, the model input data also includes external variable data that affects the actual amount of electricity bills collected; after obtaining the predicted amount of electricity bills receivable for the current month to be predicted based on the predicted electricity sales volume and the predicted electricity sales unit price, it also includes: establishing an actual amount of electricity bills collection prediction model, using the actual amount of electricity bills collection prediction model to predict the external variable data and the predicted amount of electricity bills receivable to obtain the predicted actual amount of electricity bills collected.

[0021] It can be seen from this that considering that the receivable electricity charges and the actual electricity charges collected in the same city should have a strong correlation, and the actual electricity charges collected will also be affected by external variable data such as weather and holidays, it is necessary to combine the external variable data and the predicted electricity charges receivable amount for prediction, so as to improve the accuracy of the predicted electricity charges actually collected.

[0022] In a further solution, after obtaining the predicted actual amount of electricity charges, it also includes: performing a model evaluation operation.

[0023] It can be seen that performing model evaluation operations can ensure the accuracy and effectiveness of the model.

[0024] In order to achieve the second purpose of the present invention, the present invention provides a computer device including a processor and a memory, the memory stores a computer program, and when the computer program is executed by the processor, the steps of the above-mentioned electricity fee income prediction method are implemented.

[0025] In order to achieve the third objective of the present invention, the present invention provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a controller, the steps of the above-mentioned electricity fee income prediction method are implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of an embodiment of the method for predicting electricity fee income of the present invention.

[0027] The present invention is further described below in conjunction with the accompanying drawings and embodiments. DETAILED DESCRIPTION

[0028] Electricity fee income prediction method embodiment: The electricity fee income prediction method of the present invention is a computer program applied to a computer device and is used by a power grid enterprise to predict electricity fee income.

[0029] In this embodiment, when the electricity fee revenue prediction method is working, step S1 is first executed to collect model input data of multiple dimensions through the marketing system of the power grid enterprise. Among them, the model input data includes historical electricity sales data and historical electricity sales unit price data corresponding to the transaction mode, electricity consumption category, voltage level and peak and valley. The model input data also includes external variable data that affects the actual amount of electricity charges. Through the marketing system of the power grid enterprise, data information involving multiple key dimensions is widely collected. For example, in the dimension of transaction mode, three different types of transaction modes are included, namely, guaranteed users, agent electricity purchase users, and market-based transaction users. The electricity sales situation under each mode will be the focus of analysis. In terms of electricity consumption categories, multiple categories such as large industrial electricity, non-industrial ordinary electricity, agricultural electricity, commercial electricity, residential electricity, and rice field irrigation and drainage electricity are carefully divided. The demand characteristics and consumption patterns of different electricity consumption categories are different. Comprehensive collection of these data helps to gain a deep understanding of the electricity consumption behavior of different user groups. The voltage level also covers multiple levels such as 1-10 kV, 35-110 kV, 220 kV and above. Different voltage levels correspond to electricity consumption scenarios of different scales and natures. Their electricity sales data are of great value for analyzing the load distribution and power supply efficiency of the power grid. In addition, peak and valley periods are also taken into consideration, including flat sections, valleys, peaks, peaks and other periods. The differences in electricity sales and unit prices during these periods can often reflect the time distribution of users' electricity consumption and the supply and demand of the power grid at different times. In addition to the data of the above core dimensions, full attention is also paid to the potential impact of external variable data on the electricity sales business, and holiday data and information such as the monthly minimum temperature or monthly maximum temperature are specially collected. During holidays, due to changes in the laws of people's lives and production activities, electricity demand often fluctuates significantly. The temperature factor, whether it is the monthly minimum temperature or the maximum temperature, is closely related to the frequency of use of electrical equipment by residents and enterprises. For example, in the cold winter (corresponding to the low monthly minimum temperature), the large-scale use of heating equipment will significantly increase the use of electricity. In the hot summer (corresponding to the high monthly maximum temperature), the frequent operation of refrigeration equipment will also increase the power consumption. These external variable data will serve as important supplementary information, combined with the core dimension data, to provide a more comprehensive basis for subsequent model analysis and prediction.

[0030] After obtaining the model input data of multiple dimensions, step S2 is executed to organize and clean the model input data. Since the dimensions involved in the model input data are numerous and complex, the original data may have various irregularities, inaccuracies and abnormalities. Therefore, before conducting subsequent in-depth analysis, the data must be carefully organized and cleaned to ensure the quality and availability of the data. The multi-dimensional data sources may cause large differences in data format, data magnitude and data recording methods. For example, the units for recording electricity sales under different transaction methods may not be completely consistent, or different electricity consumption categories may use different precisions when statistically analyzing voltage level-related data. If these differences are not handled, it will bring many inconveniences to subsequent data analysis and model calculations, and may even lead to erroneous results. By organizing the data, the data format can be unified and the data recording method can be standardized, so that data of different dimensions can be processed and analyzed under a relatively consistent framework.

[0031] In this implementation, the steps of arranging and cleaning the model input data include: using a preset multiple IQR (interquartile range) outlier filtering method to filter the model input data. The preset multiple can be pre-set according to the experimental data, and preferably, the preset multiple is 1.5 times. The outlier filtering method is a technique well known to those skilled in the art and will not be described in detail here. By filtering the model input data using the preset multiple IQR outlier filtering method, the influence of dirty data and abnormal data can be avoided, ensuring that the data ultimately used for analysis and prediction are relatively "clean" data within a reasonable range.

[0032] After the model input data is sorted and cleaned, step S3 is executed to distinguish the model input data to obtain the first characteristic data of the time series with trend or stability and the second characteristic data of the time series without trend or stability. Since it is not possible to guarantee that the time series of each indicator has a good trend or stability, it is necessary to distinguish the data characteristics of different indicators of the model input data so that a more appropriate prediction method can be adopted for indicators with different characteristics in the subsequent prediction link.

[0033] In this embodiment, the steps of distinguishing the model input data and obtaining the first characteristic data of the time series with trend or stability and the second characteristic data of the time series without trend or stability include: performing mean normalization processing on the model input data corresponding to each prediction indicator and calculating the coefficient of variation; if the coefficient of variation is less than or equal to the coefficient of variation fluctuation threshold, the model input data corresponding to the prediction indicator is the first characteristic data; if the coefficient of variation is greater than the coefficient of variation fluctuation threshold, the model input data corresponding to the prediction indicator is the second characteristic data. The coefficient of variation fluctuation threshold can be preset according to experimental data, and preferably, the coefficient of variation fluctuation threshold is 0.5.

[0034] Specifically, for each prediction indicator, the model input data must first be mean normalized. The purpose of mean normalization is to map the data to a relatively uniform interval so that data of different magnitudes are comparable. The specific operation is to subtract the mean of the indicator data sequence from each data point, and then divide it by the mean to obtain a new data sequence after mean normalization. Next, the coefficient of variation (CV) is calculated, and its calculation formula is the standard deviation divided by the mean. By calculating the standard deviation of the new data sequence after mean normalization and dividing it by the mean of the sequence, the coefficient of variation of the indicator can be obtained. Finally, the calculated coefficient of variation is compared with the set coefficient of variation fluctuation threshold. If the coefficient of variation is less than or equal to the coefficient of variation fluctuation threshold, it is considered that this prediction indicator has a good trend or stability, and can be confirmed as the first characteristic data; conversely, if the coefficient of variation is greater than the coefficient of variation fluctuation threshold, it is considered that this prediction indicator does not have a good trend or stability, and can be confirmed as the second characteristic data.

[0035] After distinguishing the model input data, execute step S4 to establish a general prediction model corresponding to multiple prediction indicators. After clarifying whether each prediction indicator has a trend or stability, it is necessary to establish appropriate prediction models for the two indicators with different characteristics to achieve accurate prediction of each sub-indicator. For the first characteristic data, the established general prediction model is: at least two preset time series algorithms are used to perform combined prediction on the first characteristic data to obtain the sub-model predicted electricity sales and sub-model predicted electricity unit price corresponding to the first characteristic data. For the second characteristic data, the established general prediction model is: according to the weighted addition of the indicator value of the same period last year and the indicator value of the previous month, the sub-model predicted electricity sales and sub-model predicted electricity unit price corresponding to the second characteristic data are obtained. Among them, the weighted values ​​of the indicator value of the same period last year and the indicator value of the previous month can be pre-set according to the experimental data. Preferably, the weighted values ​​of the indicator value of the same period last year and the indicator value of the previous month are both 0.5. The preset time series algorithm can be set as needed. In this embodiment, the preset time series algorithm includes ARIMA algorithm, Prophet algorithm, exponential smoothing method, gray model algorithm, autoregressive conditional heteroscedasticity model algorithm and threshold autoregressive model algorithm. When at least two preset time series algorithms are used to perform combined prediction on the first characteristic data, each algorithm can be used to predict separately, and then the best prediction value can be selected from the prediction results as the final prediction value; after each algorithm is used to predict separately, the prediction results of all algorithms are weighted and superimposed to obtain the final prediction value.

[0036] For example, for the first characteristic data with trend or stability of the subdivided index data, a combination of ARIMA and Prophet time series algorithms is used for prediction. The ARIMA algorithm is good at capturing the autocorrelation and moving average of the data, and can handle the linear relationship in the data well, but it has difficulties in handling nonlinear relationships and adding external variables. The Prophet algorithm is good at handling data with nonlinear relationships and holiday effects, but it is not sensitive enough to seasonality and trends. By combining these two algorithms, their respective advantages can be fully utilized and their respective shortcomings can be compensated, so as to achieve accurate prediction of indicator data with trend or stability. For the second characteristic data of the subdivided index data without trend or stability, a relatively simple but effective prediction method is adopted. That is, the index value of the same period last year and the index value of the previous month are used to calculate 0.5 * the index value of the same period last year + 0.5 * the index value of the previous month as the index value for the current month. The advantage of this method is that it can minimize the impact of the predicted value on the strong fluctuation of the time series, because it comprehensively considers the index values ​​of the near term and the relatively long term, and can give a relatively reasonable predicted value when the data fluctuates greatly and there is no obvious trend.

[0037] After establishing a general prediction model corresponding to multiple prediction indicators, execute step S5, use the general prediction model corresponding to each prediction indicator to predict the model input data, and obtain the sub-model predicted electricity sales volume and sub-model predicted electricity unit price corresponding to each prediction indicator of the current month to be predicted. After establishing a general prediction model for the prediction indicator, the general prediction model can be used to predict the electricity sales volume and electricity unit price data. In this embodiment, the step of predicting the model input data using the general prediction model corresponding to each prediction indicator includes: using at least two preset time series algorithms to perform a combined prediction on the first characteristic data, and obtaining the sub-model predicted electricity sales volume and sub-model predicted electricity unit price corresponding to the first characteristic data; performing weighted addition according to the indicator value of the same period last year and the indicator value of the previous month, and obtaining the sub-model predicted electricity sales volume and sub-model predicted electricity unit price corresponding to the second characteristic data.

[0038] The key to this step is to accurately input the model input data of the finest dimension under each dimension collected into the model in order to obtain accurate prediction results. The finest dimensional data of each transaction mode, electricity consumption category, voltage level, peak and valley, etc. collected from the marketing system of the power grid enterprise are accurately input into the established indicator general prediction model according to the requirements of the model. In the input process, the integrity and accuracy of the data must be ensured. Any missing or wrong data may lead to inaccurate prediction results. After the model receives the input data, it will predict the power sales volume and the unit price of power sales at the same time. Since there is a close relationship between the power sales volume and the unit price of power sales, their changes often affect each other. Therefore, synchronous prediction can more comprehensively reflect the actual situation of the power sales business of the power grid enterprise. Through the operation of the model, the predicted values ​​of the subdivided indicators of the current month can be obtained. For example, the sub-model predicted power sales volume and the sub-model predicted unit price of power sales under the guaranteed user transaction mode can be obtained.

[0039] After obtaining the sub-model predicted electricity sales and sub-model predicted electricity unit price corresponding to each prediction indicator of the current month to be predicted, execute step S6 to summarize the sub-model predicted electricity sales and sub-model predicted electricity unit price to obtain the predicted electricity sales and predicted electricity unit price of the current month to be predicted. After obtaining the sub-model predicted electricity sales and sub-model predicted electricity unit price of each subdivided indicator, these predicted values ​​need to be summarized. By summarizing the predicted values ​​of subdivided indicators under different transaction methods, electricity consumption categories, voltage levels, peaks, valleys and other dimensions, the predicted electricity sales and predicted electricity unit price of the current month to be predicted in a certain city can be obtained. This summary method can reflect the electricity sales business of the entire city from a macro perspective, and provide a basis for subsequent calculations of electricity bills and other operations.

[0040] After obtaining the predicted electricity sales volume and the predicted electricity sales unit price for the current month to be predicted, step S7 is executed to obtain the predicted electricity fee receivable amount for the current month to be predicted based on the predicted electricity sales volume and the predicted electricity sales unit price. The predicted electricity fee receivable amount for the current month to be predicted is obtained by the following formula: predicted electricity fee receivable amount for the current month to be predicted = predicted electricity sales volume for the current month to be predicted × predicted electricity sales unit price for the current month to be predicted.

[0041] After obtaining the predicted electricity fee receivable amount for the current month to be predicted, execute step S8 to establish a prediction model for the actual amount of electricity fees, and use the prediction model for the actual amount of electricity fees to predict the external variable data and the predicted electricity fee receivable amount to obtain the predicted actual amount of electricity fees. Considering that the receivable electricity fees and the actual amount of electricity fees collected in the same city should have a strong correlation, and the actual amount of electricity fees collected will also be affected by external variable data such as weather and holidays, it is necessary to combine the external variable data and the predicted electricity fee receivable amount for prediction, so as to improve the accuracy of the predicted actual amount of electricity fees collected.

[0042] In order to build a prediction model for the actual amount of electricity bills collected, the number of holidays in the month, the number of working days in the month, the number of Saturdays and Sundays in the month, the lowest temperature in the month (or the highest temperature in the month), and the amount of electricity bills to be collected in the month are selected as features. The selection of these features is based on an in-depth analysis of the factors affecting the actual amount of electricity bills collected. During holidays, people's living and production activities change, which may affect the actual collection of electricity bills; the distribution of working days and Saturdays and Sundays will also affect the actual collection of electricity bills, because people have different electricity usage habits in different time periods; and weather factors, whether it is the lowest temperature or the highest temperature of the month, are closely related to the frequency of use of electrical equipment by residents and enterprises, thus affecting the actual collection of electricity bills; the amount of electricity bills to be collected in the month itself, as an important reference value, is also closely related to the actual amount of electricity bills collected. When selecting an algorithm for predicting the actual amount of electricity bills collected in the month, the gradient boosting decision tree (XGBoost) algorithm is selected. Because the XGBoost algorithm can effectively capture linear or complex nonlinear relationships between features, and it is a centralized learning algorithm that can train weak classifiers through multiple iterations to improve the accuracy of the model. In addition, according to experimental verification, the southern region is more sensitive to the lowest temperature, and the northern region is more sensitive to the highest temperature. Therefore, when selecting weather characteristics, choosing the lowest temperature or the highest temperature of the month as the feature based on the geographical location of the city can better reflect the actual local situation.

[0043] After obtaining the predicted actual amount of electricity charges, execute step S9 to perform a model evaluation operation. By performing a model evaluation operation, the accuracy and effectiveness of the model can be ensured. In this embodiment, the mean absolute percentage error (MAPE) is used to evaluate the effect of the model. The model evaluation scope of this embodiment is all prediction indicators, including the prediction effect of subdivided indicators. By evaluating all prediction indicators, subdivided indicators with poor prediction effects can be captured, thereby providing users with a way to check whether there are abnormal events or manual errors that cause abnormal data performance. This evaluation method can help users find problems in a timely manner, adjust models or process abnormal data, thereby improving the accuracy and effectiveness of the model.

[0044] In a specific example, the model consists of five sub-models, namely, monthly electricity sales forecast model, monthly electricity bill receivable forecast model, monthly actual electricity bill collected forecast model, daily actual electricity bill collected forecast model and daily actual electricity bill collected rolling model.

[0045] The main task of the monthly electricity sales forecasting model is to predict the monthly electricity sales of each transaction type. For example, the monthly total electricity sales forecast is based on the electricity sales details table - electricity price excluding fuel historical data, and the monthly total electricity sales are predicted through a combination of models (ARIMA, Prophet, exponential smoothing, gray model, autoregressive conditional heteroscedasticity model, threshold autoregressive model). The user electricity sales model forecasts for different user categories such as proxy electricity purchases, guaranteed users, and market-based transactions are calculated separately. Taking proxy electricity purchase users as an example, combined with the monthly total electricity sales forecast value, the monthly electricity sales proportion of proxy electricity purchase users is decomposed into the corresponding month, thereby obtaining the proxy electricity sales forecast value of future months. The monthly electricity sales forecast value of proxy electricity purchase users = the proportion of total electricity sales of proxy electricity purchase users in the same period last year * the monthly total electricity sales forecast value.

[0046] The monthly electricity bill receivable forecast model first predicts the monthly electricity bill receivable in the electricity sales details based on the historical data in the electricity sales details table and the cross-reference relationship between various fees. Then, it predicts the future monthly pre-collected electricity bills and monthly pre-collected electricity bills based on the historical values ​​of pre-collected electricity bills and pre-collected electricity bills. Finally, the final monthly electricity bill receivable forecast value is calculated from the three.

[0047] For example, taking the monthly agent purchase electronic model as an example: (1) Monthly forecast of transmission and distribution charges for power purchasing agents: The first step is to obtain the transmission and distribution unit price based on the electricity consumption classification in the provincial transmission and distribution price catalogue published by the state; the second step is to first use the transmission and distribution unit price output in the first step and the monthly electricity sales forecast to calculate and obtain the transmission and distribution fee for the electricity consumption classification at the peak and valley levels; then the transmission and distribution fees are gradually summarized to obtain the transmission and distribution fees at the electricity consumption category level.

[0048] (2) Monthly proxy electricity purchase price forecast for proxy electricity purchase users: The first step is to obtain the proxy electricity purchase unit price based on the electricity consumption classification in the provincial electricity price catalogue published by the state; the second step is to first use the proxy electricity purchase unit price output from the first step and the monthly electricity sales forecast value to calculate and obtain the proxy electricity purchase price at the peak and valley levels; then the proxy electricity purchase prices are gradually summarized to obtain the proxy electricity purchase prices at the electricity consumption category level.

[0049] (3) Monthly power factor adjustment electricity fee forecast for power purchasing users: The power factor adjustment electricity fee historical value is predicted based on historical data through a combination of models (ARIMA, exponential smoothing, grey model, autoregressive conditional heteroscedasticity model, threshold autoregressive model).

[0050] (4) Monthly forecast of basic electricity charges for power purchasing agents: The first step is to directly obtain the unit price of the corresponding transformer capacity (kVA) and the unit price of the maximum demand (kW); the second step is to predict the transformer capacity (kVA) and the maximum demand (kW) based on historical data through a combination of models (ARIMA, exponential smoothing, grey model, autoregressive conditional heteroscedasticity model, threshold autoregressive model); the third step is to obtain the maximum capacity fee and maximum demand fee according to the following calculation formula: maximum capacity fee = transformer capacity (kVA) * unit price of transformer capacity (kVA), maximum demand fee = maximum demand (kW) * unit price of maximum demand (kW); the fourth step is to obtain the basic electricity charge forecast value according to the following calculation formula: basic electricity charge forecast value = maximum capacity fee + maximum demand fee; (5) Monthly forecast of the fund and surcharge for power purchasing agents: The first step is to use a combination of models (ARIMA, exponential smoothing, grey model, autoregressive conditional heteroscedasticity model, threshold autoregressive model) to forecast the unit price of the fund and surcharge based on the power sales schedule - historical data of power purchasing agents. The second step is to use the forecasted unit price of the fund and surcharge to calculate the fund and surcharge with the forecasted monthly power sales volume.

[0051] (6) Forecast of monthly electricity charges receivable from power purchasing agents: The monthly electricity charges receivable are calculated based on the cross-references between transmission and distribution charges, power purchasing charges, power factor adjustment charges, basic charges, funds and surcharges.

[0052] The forecast value of electricity charges that should be receivable by power purchasing agents this month = transmission and distribution charges + power purchasing charges for agents + power factor adjustment charges + basic charges + funds and surcharges.

[0053] The monthly electricity bill forecasting model constructs a monthly electricity bill calculation equation and predicts the monthly electricity bill by analyzing the monthly electricity bill receivable and other historical data that may affect the monthly electricity bill (user payment habits, holiday factors, etc.).

[0054] The daily actual electricity charge forecasting model decomposes the daily actual electricity charge based on the results of the monthly actual electricity charge forecasting model. The daily predicted actual electricity charge can be obtained by multiplying the proportion of historical records in the same period of the previous year with the results of the monthly actual electricity charge forecasting model.

[0055] The daily actual electricity bill rolling model goes a step further by using the actual electricity bill values ​​known for the current month to update the forecast values ​​for future dates.

[0056] It should be noted that the monthly electricity sales forecast model, the monthly electricity receivable forecast model, and the monthly actual electricity fee forecast model are the main models, and the daily actual electricity fee forecast model and the daily actual electricity fee rolling model are used for auxiliary supplements.

[0057] From the above, it can be seen that the electricity fee revenue prediction method of the present invention collects model input data of multiple dimensions, uses the general prediction model corresponding to multiple prediction indicators to predict the model input data, and summarizes the predicted sub-model predicted electricity sales volume and sub-model predicted electricity unit price to obtain the predicted electricity sales volume and predicted electricity unit price of the current month to be predicted, and then obtains the electricity fee receivable amount of the current month to be predicted. It can realize multi-dimensional and multi-category analysis in combination with actual business needs. The application scope of the model calculation results is relatively wide, and more fine-grained control can be performed from the dimensions of electricity consumption category, peak and valley, and voltage level, thereby improving the accuracy and stability of the prediction.

[0058] Computer device embodiment: The computer device of this embodiment includes a controller, and when the controller executes a computer program, the steps in the above-mentioned electricity fee income prediction method embodiment are implemented.

[0059] For example, a computer program may be divided into one or more modules, one or more modules are stored in a memory and executed by a controller to implement the present invention. One or more modules may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program in a computer device.

[0060] The computer device may include, but is not limited to, a controller and a memory. Those skilled in the art will appreciate that the computer device may include more or fewer components, or a combination of certain components, or different components, for example, the computer device may also include input and output devices, network access devices, buses, etc.

[0061] For example, the controller may be a central processing unit (CPU), other general controllers, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general controller may be a microcontroller or any conventional controller, etc. The controller is the control center of the computer device, and uses various interfaces and lines to connect the various parts of the entire computer device.

[0062] The memory can be used to store computer programs and / or modules. The controller realizes various functions of the computer device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. For example, the memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound receiving function, a sound conversion to text function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, text data, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SmartMediaCard, SMC), a secure digital (SecureDigital, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0063] Computer readable storage medium embodiment: If the module integrated in the computer device of the above embodiment is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the implementation of all or part of the process in the above electricity fee revenue prediction method embodiment can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the controller, the steps of the above electricity fee revenue prediction method embodiment can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The storage medium may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electric carrier signals and telecommunication signals.

[0064] It should be noted that the above are only preferred embodiments of the present invention, but the design concept of the invention is not limited thereto, and any non-substantial modifications made to the present invention using this concept also fall within the protection scope of the present invention.

Claims

1. A method for predicting electricity bill income, characterized in that: include: Collecting model input data of multiple dimensions through the marketing system of power grid enterprises, the model input data including transaction mode, electricity consumption category, voltage level, and historical electricity sales data and historical electricity sales price data corresponding to peak and valley levels; Establish a general prediction model corresponding to multiple prediction indicators; The model input data is predicted using the general prediction model corresponding to each prediction indicator to obtain the sub-model predicted electricity sales volume and the sub-model predicted electricity sales unit price corresponding to each prediction indicator of the current month to be predicted; Summarize the predicted electricity sales volume and the predicted electricity selling unit price of the sub-model to obtain the predicted electricity sales volume and the predicted electricity selling unit price of the current month to be predicted; The predicted amount of electricity charges to be receivable for the current month to be predicted is obtained based on the predicted electricity sales volume and the predicted electricity sales unit price.

2. The electricity fee income prediction method according to claim 1, characterized in that: After collecting model input data of various dimensions through the power grid enterprise marketing system, it also includes: The model input data is differentiated to obtain first characteristic data of a time series with trend or stability and second characteristic data of a time series without trend or stability.

3. The electricity fee income prediction method according to claim 2, characterized in that: The step of distinguishing the model input data to obtain first characteristic data of a time series having a trend or stability and second characteristic data of a time series not having a trend or stability comprises: Performing mean normalization processing on the model input data corresponding to each of the prediction indicators, and calculating the coefficient of variation; If the coefficient of variation is less than or equal to the coefficient of variation fluctuation threshold, the model input data corresponding to the prediction index is the first characteristic data; If the coefficient of variation is greater than the coefficient of variation fluctuation threshold, the model input data corresponding to the prediction index is the second characteristic data.

4. The electricity fee income prediction method according to claim 2, characterized in that: The step of using the general prediction model corresponding to each prediction indicator to predict the model input data includes: Using at least two preset time series algorithms to perform combined prediction on the first characteristic data to obtain a sub-model predicted electricity sales volume and a sub-model predicted electricity sales unit price corresponding to the first characteristic data; and / or The sub-model predicted electricity sales amount and the sub-model predicted electricity sales unit price corresponding to the second characteristic data are obtained by performing weighted addition on the index value of the same period last year and the index value of the previous month.

5. The electricity fee income prediction method according to claim 1, characterized in that: After the steps of collecting model input data of various dimensions through the power grid enterprise marketing system, it also includes: The model input data is sorted and cleaned.

6. The electricity fee income prediction method according to claim 5, characterized in that: The steps of arranging and cleaning the model input data include: The model input data is filtered using an outlier filtering method with a preset multiple of IQR.

7. The electricity fee income prediction method according to any one of claims 1 to 6, characterized in that: The model input data also includes external variable data that affects the actual amount of electricity charges; After obtaining the predicted amount of electricity charges to be receivable for the current month to be predicted according to the predicted electricity sales volume and the predicted electricity sales unit price, the method further includes: Establish a prediction model for the actual amount of electricity charges; The electricity fee actual collection amount prediction model is used to predict the external variable data and the predicted electricity fee receivable amount to obtain the predicted electricity fee actual collection amount.

8. The electricity fee income prediction method according to claim 7, characterized in that: After obtaining the predicted electricity bill actual amount, it also includes: Perform model evaluation operations.

9. A computer device comprising a processor and a memory, characterized in that: The memory stores a computer program, and when the computer program is executed by the processor, the steps of the electricity fee income prediction method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the controller, the steps of the electricity rate revenue prediction method according to any one of claims 1 to 8 are implemented.

Images