Regional electricity consumption intelligent prediction method based on multi-source heterogeneous data

Abstract

The invention relates to the technical field of electricity consumption prediction, in particular to an intelligent regional electricity consumption prediction method based on multi-source heterogeneous data, which comprises the following steps: determining an electricity consumption prediction influence tendency for a next preset monitoring period based on an electric power change characterization value; when it is determined that the next preset monitoring period is the power consumption prediction strong influence tendency, it is determined that the power consumption is corrected through the elasticity influence weight coefficient, and whether prediction for the power consumption is qualified or not is determined based on the energy change characterization value; and when it is determined that prediction for the electricity consumption is abnormal, correcting prediction parameters for determining the electricity consumption in the next preset monitoring period based on a regional industry change quantized value, including identifying a main analysis data set, determining to add a model training set or separate a prediction model based on a periodic quantized value of the main analysis data, and determining to determine the electricity consumption in the next preset monitoring period. According to the specific condition of the power elastic coefficient of the to-be-predicted region, the power consumption is analyzed and predicted in a targeted manner, and the prediction efficiency of the power consumption in the region is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power consumption prediction, in particular to a regional power consumption intelligent prediction method based on multi-source heterogeneous data. BACKGROUND

[0002] In today's society, electricity as a vital energy, its stable supply and reasonable distribution play a key role in economic development, social stability and people's quality of life. Accurate prediction of regional power consumption not only helps power companies to reasonably arrange power generation plans, optimize power grid dispatching, reduce power generation costs and reduce energy waste, but also provides a scientific basis for the government to formulate energy policies and plan energy infrastructure construction.

[0003] At present, there are many methods for predicting regional power consumption, but with the rapid development of economy, continuous adjustment of energy structure and increasing diversification of user power consumption behavior, traditional prediction methods often fail to fully consider the influence of multi-source heterogeneous data, resulting in certain limitations in the accuracy and reliability of the prediction results. Multi-source heterogeneous data is widely sourced and has various formats, and contains rich information related to power consumption.

[0004] In the city, with the improvement of residents' living standards and the continuous improvement of urban functions, residential electricity and commercial electricity demand show different trends; in industrial parks, the upgrading and adjustment of industrial structure will directly affect the change of industrial electricity consumption. Economic activities are complex and diverse, energy consumption structure is constantly changing, and industrial development is in a dynamic adjustment process, which poses great challenges to power consumption prediction.

[0005] Chinese patent publication No. CN116108979A discloses a cross-cycle multi-source heterogeneous power data processing system, which includes a data acquisition module, a data analysis module, a data modeling module, a multi-source heterogeneous information module, a prediction module, a switching module, and a data pushing module. The data acquisition module, the data analysis module, the data modeling module, the multi-source heterogeneous information module, the prediction module, the switching module, and the data pushing module are application software installed in a PC. The application method of the cross-cycle multi-source heterogeneous power data processing system includes seven steps. The power consumption area power load data model can be obtained, and the power consumption can be predicted according to the obtained model. The power in the area can be adjusted in advance before the power changes. It can be seen that the above technical solution has the following problems: it does not consider the specific situation of the power elasticity coefficient of the area to be predicted, and the power consumption is analyzed and predicted, which affects the prediction efficiency of the power consumption in the area. SUMMARY

[0006] To this end, the application provides a regional power consumption intelligent prediction method based on multi-source heterogeneous data, to overcome the problem that the power consumption is not analyzed and predicted in a targeted manner according to the specific situation of the power elasticity coefficient of the region to be predicted, affecting the prediction accuracy of the power consumption in the region.

[0007] To achieve the above-mentioned purpose, the application provides a regional power consumption intelligent prediction method based on multi-source heterogeneous data, comprising: collecting the power consumption in each historical preset monitoring period in the region to construct a power consumption dataset; determining a power change representation value based on the historical power elasticity coefficient; determining the power consumption prediction influence tendency for the next preset monitoring period based on the power change representation value; when it is determined that the next preset monitoring period is a strong influence tendency for power consumption prediction, correcting the power consumption by an elasticity influence weight coefficient, and determining whether the prediction for the power consumption is qualified based on the energy change representation value; when it is determined that the prediction for the power consumption is abnormal, correcting the prediction parameter for the power consumption in the next preset monitoring period based on the regional industry change quantitative value, including identifying a main analysis dataset, determining to increase a model training set or separate a prediction model based on the period quantitative value of the main analysis data; when it is determined that the next preset monitoring period is a weak influence tendency for power consumption prediction, determining the power consumption in the next preset monitoring period based on the power consumption dataset.

[0008] Further, the process of determining a power change representation value based on a historical power elasticity coefficient comprises: drawing a power elasticity coefficient time domain curve based on each power elasticity coefficient in the historical data; solving the absolute value of the ratio of the slope absolute value of the power elasticity coefficient time domain curve at the current time node to the preset stable slope to obtain the power change representation value.

[0009] Further, the process of determining the power consumption prediction influence tendency for the next preset monitoring period based on the power change representation value comprises: when the power change representation value is less than or equal to a preset power change representation value, determining that the next preset monitoring period is a weak influence tendency for power consumption prediction, and determining the power consumption in the next preset monitoring period based on the power consumption dataset; when the power change representation value is greater than the preset power change representation value, determining that the next preset monitoring period is a strong influence tendency for power consumption prediction, and correcting the power consumption by an elasticity influence weight coefficient.

[0010] Further, the process of correcting the power consumption by an elasticity influence weight coefficient comprises: Determine the elastic coefficient selection interval based on the current power elastic coefficient; Determine each preset monitoring period in which the power elastic coefficient in the historical data is within the elastic coefficient selection interval as a reference period; Obtain the influence multiple of each reference period; Determine the elastic influence weight coefficient based on each influence multiple.

[0011] Further, the process of determining the energy change representation value includes: Obtain the ratio of the natural gas consumption in the recent preset monitoring period to the average natural gas consumption in each preset monitoring period in the historical data to obtain the natural gas change rate; Obtain the ratio of the coal consumption in the recent preset monitoring period to the average coal consumption in each preset monitoring period in the historical data to obtain the coal change rate; Obtain the ratio of the oil consumption in the recent preset monitoring period to the average oil consumption in each preset monitoring period in the historical data to obtain the oil change rate; Obtain the ratio of the electricity consumption in the recent preset monitoring period to the average electricity consumption in each preset monitoring period in the historical data to obtain the electricity change rate; Calculate the average of the natural gas change rate, the coal change rate, and the oil change rate to obtain the alternative energy change amount; Solve the absolute value of the difference between the alternative energy change amount and the electricity change rate and the ratio of the electricity change rate to obtain the energy change representation value.

[0012] Further, based on the energy change representation value, determine whether the prediction for electricity consumption is qualified, including: When the energy change representation value is less than or equal to the preset energy change representation value, it is determined that the prediction for electricity consumption is qualified, and the product of the elastic influence weight coefficient and the electricity consumption predicted by the first prediction model is determined as the predicted electricity consumption in the next preset monitoring period; When the energy change representation value is greater than the preset energy change representation value, it is determined that the prediction for electricity consumption is abnormal, and the prediction parameter used to determine the electricity consumption in the next preset monitoring period is corrected based on the regional industrial change quantitative value.

[0013] Further, the process of determining the regional industrial change quantitative value includes: Obtain the industrial electricity consumption and service industry electricity consumption in each historical preset monitoring period in the region to obtain the industrial electricity data set and the service electricity data set, respectively; Obtain the ratio of the industrial electricity consumption in the recent preset monitoring period to the average industrial electricity consumption in each historical preset monitoring period to obtain the regional industrial change quantitative value; The ratio of the service electricity consumption of the latest preset monitoring period to the average service electricity consumption of each historical preset monitoring period is obtained to obtain the regional service industry change quantitative value.

[0014] Further, the process of correcting the prediction parameter used to determine the electricity consumption in the next preset monitoring period based on the regional industry change quantitative value includes: When the regional industrial industry change quantitative value is less than or equal to the regional service industry change quantitative value, the service electricity consumption dataset is determined as the main analysis dataset; When the regional industrial industry change quantitative value is greater than the regional service industry change quantitative value, the industrial electricity consumption dataset is determined as the main analysis dataset.

[0015] Further, the process of determining an increased model training set or separating a prediction model based on the cycle quantitative value of the main analysis data includes: Based on the obtained electricity consumption of each preset monitoring period in the main analysis dataset, a main electricity consumption time domain curve is drawn; The signal of the main electricity consumption time domain curve is converted into a frequency domain signal through Fourier transform to obtain a frequency spectrum graph; Each peak value in the frequency spectrum graph is obtained; When there is a peak value greater than a preset peak value, it is determined that the model training set is increased; When each peak value is less than or equal to the preset peak value, the prediction model is separated.

[0016] Further, the process of increasing the model training set includes: Based on the maximum peak value, the data amount of the training set used to train the first prediction model is adjusted to a corresponding value, wherein, Each peak value in the frequency spectrum graph is obtained to obtain the maximum peak value; The increase amplitude of the data amount is negatively correlated with the maximum peak value; The process of separating the prediction model includes: The predicted electricity consumption output by the model trained through the industrial electricity consumption dataset is determined as the industrial predicted electricity consumption; The predicted electricity consumption output by the model trained through the service electricity consumption dataset is determined as the service predicted electricity consumption; The sum of the industrial predicted electricity consumption and the service predicted electricity consumption is determined as the predicted electricity consumption of the next preset monitoring period.

[0017] Compared with the prior art, the present application has the beneficial effect that the power change characteristic value is used to determine the power consumption prediction influence tendency for the next preset monitoring period, and the power change characteristic value reflects the change degree of the power elasticity coefficient. When the power change characteristic value is less than or equal to the preset power change characteristic value, the change rate is similar to the stable state, and the economic and power consumption relationship is stable; when the power change characteristic value is greater than the preset power change characteristic value, the change is violent, and at this time, the economic and power consumption relationship fluctuates, which affects the prediction accuracy, and the internal law between the economic relationship and the power consumption changes rapidly, and only using the model trained based on the past stable law to determine the prediction value has high risk. Before directly using the power consumption data for prediction, the elasticity coefficient is used to judge the actual situation of the prediction environment, and a decision basis is provided for determining whether to start the complex correction process, which effectively saves the operation resources and further improves the prediction efficiency of the power consumption.

[0018] Further, when it is judged that the influence tendency is strong, a period with similar elasticity coefficients is identified from historical data as a reference, and the elasticity influence weight coefficient represents the corresponding relationship between the historical prediction value and the true value in the similar economic environment as the current one, which is an empirical correction factor. The prediction error in the similar historical scenario is used to correct the current prediction, so that the prediction result is fused with historical experience, which improves the prediction accuracy of the power consumption and further improves the prediction efficiency of the power consumption.

[0019] Further, the energy change characteristic value is used for prediction qualification test, and the energy change characteristic value measures the consistency between the alternative energy change and the power consumption change. When the energy change characteristic value is greater than the preset energy change characteristic value, the power consumption behavior is different from other energies, and energy substitution will affect the power consumption prediction. The prediction parameters for determining the power consumption in the next preset monitoring period are corrected based on the regional industry change quantitative value. The power consumption prediction is verified by multiple source data, and a more comprehensive evaluation system is formed. The risks caused by the internal structure change of energy, which cannot be revealed by simple power consumption data or economic data, are detected sharply, and the prediction efficiency of the power consumption is further improved.

[0020] Further, when the energy change characteristic value is greater than the preset energy change characteristic value, the energy consumption trend exists abnormality, and the industrial structure level appears change. The industrial change quantitative value quantifies the amplitude of the switching of the regional industrial gravity center between industry and service industry. When the industrial tendency of a region changes, the electricity consumption trend will fluctuate, and the abnormal situation is identified for targeted analysis. Further, the industrial change quantitative value is used to determine the main analysis data set of the changed industry for main analysis, the cycle quantitative value based on the main analysis data is used to determine the increase of the model training set or the separation of the prediction model, the influence of the industrial structure change on the electricity consumption is analyzed, and the prediction accuracy is improved. Focusing on the changed industry, the change trend of the electricity consumption is better captured, and the prediction efficiency of the electricity consumption is further improved.

[0021] Further, when the peak value in the spectrum diagram is greater than the preset peak value, the periodic fluctuation of the electricity consumption is relatively strong, and there are some complex change patterns. At this time, the model training set is increased to enable the model to learn more features and rules, and the prediction accuracy is improved. When each peak value is less than or equal to the preset peak value, the periodic fluctuation of the electricity consumption is relatively weak, and the electricity consumption characteristics of different industries are relatively independent. At this time, the prediction model is separated to make more accurate prediction for different industries. According to the actual periodic fluctuation of the electricity consumption, the processing mode is flexibly adjusted, so that the model can maintain good performance in different situations, enhance the adaptability of the model, and further improve the prediction efficiency of the electricity consumption.

[0022] Further, the cycle quantitative value based on the main analysis data is used to determine the increase of the model training set or the separation of the prediction model, and two different strategies are adopted according to the periodic characteristics of the main industrial electricity consumption data. The maximum peak value represents the intensity of the most important periodic component in the electricity consumption change. When the maximum peak value is small, the periodic fluctuation of the electricity consumption is weak, and the model needs more data to learn this complex change pattern. Therefore, the increase of the data amount is negatively correlated with the maximum peak value, that is, the greater the maximum peak value, the relatively smaller the amount of increased data, because the model can capture the main change characteristics from the existing data at this time; the smaller the maximum peak value, the relatively larger the amount of increased data, to help the model learn more detailed features. By adjusting the training set data amount according to the maximum peak value, the learning effect of the model is ensured, and the problems of long training time and overfitting caused by too large data amount are avoided, and the prediction efficiency of the electricity consumption is further improved.

[0023] Further, the electricity consumption characteristics of the industry and service industry are different, and the change of the electricity consumption is affected by different factors. When the periodic fluctuation of the electricity consumption is relatively weak, the model is trained respectively by using the industrial electricity consumption data set and the service electricity consumption data set, so that the electricity consumption law of different industries can be captured more accurately. Finally, the prediction results of the two models are added to obtain the predicted electricity consumption in the next preset monitoring period. The electricity consumption characteristics of different industries are fully considered, the mutual interference of the electricity consumption characteristics of different industries is avoided, and the prediction accuracy is improved. By separating the prediction model, the electricity consumption characteristics of the industry and service industry are modeled and predicted, the electricity consumption law of different industries is captured more accurately, more accurate basis is provided for power supply and management, and the prediction efficiency of the electricity consumption is further improved. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 The step flow chart of the regional electricity consumption intelligent prediction method based on multi-source heterogeneous data of the embodiment of the application is shown in the figure. Figure 2 The logic decision diagram for determining the electricity consumption prediction influence tendency for the next preset monitoring period based on the power change representation value of the embodiment of the application is shown in the figure. Figure 3 The logic decision diagram for determining whether the prediction of the electricity consumption is qualified based on the energy change representation value of the embodiment of the application is shown in the figure. Figure 4 The logic decision diagram for determining whether to increase the model training set or separate the prediction model based on the periodic quantization value of the main analysis data of the embodiment of the application is shown in the figure. DETAILED DESCRIPTION

[0025] In order to make the objects and advantages of the present application clearer, the present application will be further described below with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and do not limit the protection scope of the present application.

[0026] The preferred embodiments of the present application will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present application, and are not intended to limit the protection scope of the present application.

[0027] It should be noted that in the description of the present application, the terms "up", "down", "left", "right", "in", "out" and the like indicate the direction or positional relationship terms based on the direction or positional relationship shown in the drawings, which are only for the convenience of description, and do not indicate or imply that the device or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application.

[0028] Moreover, it needs to be explained that, in the description of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connecting" should be understood in a broad sense, for example, it can be fixed connection, or detachable connection, or integrally connected, it can be mechanical connection, or electrical connection, it can be directly connected, or indirectly connected through intermediate medium, or the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0029] Please refer to Figure 1 The steps of the method are shown in the flow chart of the embodiment of the present application based on multi-source heterogeneous data regional power consumption intelligent prediction method, the method comprises: S1, collecting the power consumption in each historical preset monitoring period in the region to construct a power consumption dataset; S2, determining the power change representation value based on the historical power elasticity coefficient; S3, determining the power consumption prediction influence tendency for the next preset monitoring period based on the power change representation value; S4, when it is determined that the next preset monitoring period is a strong influence tendency for power consumption prediction, the power consumption is corrected by the elasticity influence weight coefficient, and whether the prediction for power consumption is qualified is determined based on the energy change representation value; When it is determined that the prediction for power consumption is abnormal, the prediction parameters for determining the power consumption in the next preset monitoring period are corrected based on the regional industry change quantitative value, including identifying the main analysis dataset, determining the increase model training set or separating the prediction model based on the period quantitative value of the main analysis data; When it is determined that the next preset monitoring period is a weak influence tendency for power consumption prediction, the power consumption in the next preset monitoring period is determined based on the power consumption dataset.

[0030] Specifically, the process of determining the power change representation value based on the historical power elasticity coefficient comprises: Drawing the power elasticity coefficient time domain curve based on each power elasticity coefficient in the historical data; Solving the absolute value of the ratio of the slope absolute value of the power elasticity coefficient time domain curve at the current time node to the preset stable slope, to obtain the power change representation value.

[0031] Specifically, the process of determining the preset stable slope comprises: Marking the slope absolute value of the preset monitoring period corresponding to the prediction deviation value in the historical data less than or equal to the preset prediction deviation value as the stable slope; Solving the average value of each stable slope to obtain the preset stable slope; The difference absolute value between the electricity consumption predicted by the first prediction model and the actual electricity consumption is calculated, and the ratio of the difference absolute value to the actual electricity consumption is solved to obtain a prediction deviation amount; The preset prediction deviation amount is selected in the interval [0.04, 0.08], and those skilled in the art can determine the preset prediction deviation amount according to the actual application scene. It can be understood that the accuracy of the prediction of the first prediction model can be divided, and in the embodiment, the preset prediction deviation amount is preferably 0.06.

[0032] Specifically, the electricity elasticity coefficient in a single preset monitoring period is the ratio of the annual average growth rate of electricity consumption to the annual average growth rate of the national economy. Please refer to Figure 2 As shown in the figure, it is a logic decision diagram for determining the electricity consumption prediction influence tendency for the next preset monitoring period based on the power change representation value, and the process of determining the electricity consumption prediction influence tendency for the next preset monitoring period based on the power change representation value comprises: When the power change representation value is less than or equal to the preset power change representation value, it is determined that the next preset monitoring period is weakly influenced by electricity consumption prediction, and the electricity consumption in the next preset monitoring period is determined based on the electricity consumption data set. When the power change representation value is greater than the preset power change representation value, it is determined that the next preset monitoring period is strongly influenced by electricity consumption prediction, and the electricity consumption is corrected by the elasticity influence weight coefficient.

[0033] Specifically, the preset power change representation value is selected in the interval [1.07, 1.13], and those skilled in the art can select and determine the preset power change representation value according to the actual situation. The preset power change representation value can be selected by statistical analysis of historical data. When the power change representation value exceeds different thresholds in the historical data, the threshold that significantly increases the prediction deviation amount is selected as the preset power change representation value. In the embodiment, the preset power change representation value is preferably 1.07.

[0034] Specifically, the power change characteristic value is used to determine the electricity consumption prediction influence tendency for the next preset monitoring period. The power change characteristic value reflects the change degree of the power elasticity coefficient. When the power change characteristic value is less than or equal to a preset power change characteristic value, the change rate is similar to the stable state, and the relationship between the economy and the electricity consumption is stable. When the power change characteristic value is greater than the preset power change characteristic value, the change is violent, and at this time, the relationship between the economy and the electricity consumption fluctuates, which affects the prediction accuracy. The internal law between the electricity consumption and the economy changes rapidly, and the use of the model trained based on the past stable law to determine the prediction value is risky. Before directly using the electricity consumption data for prediction, the elasticity coefficient is used to judge the actual situation of the prediction environment, which provides a decision basis for determining whether to start the complex correction process, effectively saves the operation resources, and further improves the prediction efficiency of the electricity consumption.

[0035] Specifically, the first prediction model is a time series prediction model LSTM, and the specific process of training the first prediction model is not limited, which can include: The electricity consumption data set is preprocessed, including cleaning, missing value processing and abnormal value processing; The number of network layers, the number of neurons and the learning rate are set; The historical electricity consumption data is divided into a training set and a test set in chronological order, the training set is used to fit the model, and the test set is used to evaluate the performance, which is a prior art and will not be described in detail.

[0036] In a single embodiment, the number of network layers is 2, the number of neurons is 50 per layer, and the learning rate is 0.001.

[0037] Specifically, the process of determining the electricity consumption in the next preset monitoring period based on the electricity consumption data set includes determining the electricity consumption in the next preset monitoring period by the first prediction model.

[0038] Specifically, the process of correcting the electricity consumption by the elasticity influence weight coefficient includes: The elasticity coefficient selection interval is determined based on the current power elasticity coefficient; The historical data in the elasticity coefficient selection interval is determined as a reference period; The influence multiple of each reference period is obtained; The elasticity influence weight coefficient is determined based on the influence multiple.

[0039] Specifically, D0±△D is determined as the elasticity coefficient selection interval, D0 is the current power elasticity coefficient, and △D is a preset tolerance amount.

[0040] Specifically, the influence multiple of a single reference period is the ratio of the actual electricity consumption in the period to the electricity consumption predicted by the first prediction model.

[0041] Specifically, the process of determining the elasticity influence weight coefficient based on each influence multiple includes determining the median of each influence multiple as the elasticity influence weight coefficient.

[0042] Specifically, the preset tolerance amount is selected within the interval [0.09T0, 0.14T0], T0 is the average of historical power elasticity coefficients, and a person skilled in the art can select and determine the preset tolerance amount according to the historical data dispersion. The greater the fluctuation of the historical power elasticity coefficients, the greater the preset tolerance amount. In the embodiment, preferably, the preset tolerance amount is 0.1T0.

[0043] Specifically, when it is judged that the strong influence tendency, a period with similar elasticity coefficients is identified from historical data as a reference. The elasticity influence weight coefficient represents the corresponding relationship between the historical prediction value and the true value in a similar economic environment as the current one, which is an empirical correction factor. The prediction error experience in the similar scenario is used to correct the current prediction, so that the prediction result integrates the historical experience, improves the prediction accuracy of the power consumption, and further improves the prediction efficiency of the power consumption.

[0044] Specifically, the process of determining the energy change representation value includes: obtaining the ratio of the natural gas consumption in the recent preset monitoring period to the average natural gas consumption in each preset monitoring period in the historical data to obtain a natural gas change rate; obtaining the ratio of the coal consumption in the recent preset monitoring period to the average coal consumption in each preset monitoring period in the historical data to obtain a coal change rate; obtaining the ratio of the oil consumption in the recent preset monitoring period to the average oil consumption in each preset monitoring period in the historical data to obtain an oil change rate; obtaining the ratio of the power consumption in the recent preset monitoring period to the average power consumption in each preset monitoring period in the historical data to obtain a power consumption change rate; calculating the average of the natural gas change rate, the coal change rate, and the oil change rate to obtain a replaceable energy change amount; solving the ratio of the absolute value of the difference between the replaceable energy change amount and the power consumption change rate to the power consumption change rate to obtain an energy change representation value.

[0045] Specifically, the unit of natural gas consumption is ten thousand cubic meters, the unit of coal consumption is ten thousand tons, and the unit of oil consumption is ten thousand tons.

[0046] Please refer to Figure 3As shown, it is the logic decision diagram for determining whether the prediction of the power consumption is qualified based on the energy change characteristic value, the process for determining whether the prediction of the power consumption is qualified based on the energy change characteristic value comprises: When the energy change characteristic value is less than or equal to the preset energy change characteristic value, it is determined that the prediction of the power consumption is qualified, and the product of the elastic influence weight coefficient and the power consumption predicted by the first prediction model is determined as the predicted power consumption in the next preset monitoring period; When the energy change characteristic value is greater than the preset energy change characteristic value, it is determined that the prediction of the power consumption is abnormal, and the prediction parameter used to determine the power consumption in the next preset monitoring period is corrected based on the regional industrial change quantitative value.

[0047] Specifically, the preset energy change characteristic value is selected in the interval [0.05, 0.15], and the person skilled in the art can select the preset energy change characteristic value by himself, which can be determined by analyzing the historical data. It can be understood that the case of whether there is a significant change in the energy structure can be divided. In the embodiment, preferably, the preset energy change characteristic value is 0.1.

[0048] Specifically, the prediction qualification test is performed based on the energy change characteristic value, and the energy change characteristic value reflects the consistency of the change of the alternative energy and the change of the power consumption. When the energy change characteristic value is greater than the preset energy change characteristic value, the power consumption behavior is different from other energies, and the energy substitution will affect the power consumption prediction. Based on this, the prediction parameter used to determine the power consumption in the next preset monitoring period is corrected based on the regional industrial change quantitative value. Through the secondary verification of the power consumption prediction by the multi-source data, a more comprehensive evaluation system is formed. The risk caused by the change of the internal structure of the energy which cannot be revealed by the simple power consumption data or economic data is detected sharply, and the prediction efficiency of the power consumption is further improved.

[0049] Specifically, the process for determining the regional industrial change quantitative value comprises: Obtain the industrial power consumption and the service industry power consumption in each historical preset monitoring period in the region to obtain the industrial power consumption dataset and the service power consumption dataset, respectively; Obtain the ratio of the industrial power consumption in the recent preset monitoring period to the average industrial power consumption in each historical preset monitoring period to obtain the regional industrial change quantitative value; Obtain the ratio of the service power consumption in the recent preset monitoring period to the average service power consumption in each historical preset monitoring period to obtain the regional service industry change quantitative value.

[0050] Specifically, the process for correcting the prediction parameter used to determine the power consumption in the next preset monitoring period based on the regional industrial change quantitative value comprises: When the quantitative value of the regional industrial industry change is less than or equal to the quantitative value of the regional service industry change, the service power consumption dataset is determined as the main analysis dataset; When the quantitative value of the regional industrial industry change is greater than the quantitative value of the regional service industry change, the industrial power consumption dataset is determined as the main analysis dataset.

[0051] Referring to Figure 4 As shown in the figure, it is the logic decision diagram of the embodiment of the application for determining the increase of the model training set or separating the prediction model based on the cycle quantitative value of the main analysis data. The process of the application for determining the increase of the model training set or separating the prediction model based on the cycle quantitative value of the main analysis data comprises: Drawing a main power consumption time domain curve based on the power consumption of each preset monitoring cycle in the obtained main analysis dataset; Converting the signal of the main power consumption time domain curve into a frequency domain signal by Fourier transform to obtain a frequency spectrum diagram; Obtaining each peak value in the frequency spectrum diagram; When there is a peak value greater than the preset peak value, it is determined that the model training set is increased; When each peak value is less than or equal to the preset peak value, the prediction model is separated.

[0052] In a single embodiment, the preset peak value is twice the average value of the spectrum.

[0053] Specifically, when the energy change representation value is greater than the preset energy change representation value, the energy consumption trend exists abnormality, and the industrial structure level appears change. The industrial change quantitative value quantifies the amplitude of the switching of the regional industrial center between industry and service industry. When the industrial tendency of a region changes, its power consumption trend will fluctuate, and the abnormal situation is identified for targeted analysis. Further, the industry with greater change is determined as the main analysis dataset for main analysis through the industrial change quantitative value, the increase of the model training set or the separation of the prediction model is determined based on the cycle quantitative value of the main analysis data, the influence of the industrial structure change on the power consumption is analyzed, and the prediction accuracy is improved. Focus on the industry with greater change to better capture the change trend of power consumption, and further improve the prediction efficiency of power consumption.

[0054] Specifically, when there is a peak value greater than the preset peak value in the spectrum diagram, the periodic fluctuation of the power consumption is relatively strong, and there is some complex change mode, at this time, the model training set is increased to let the model learn more features and rules, and the prediction accuracy is improved. When each peak value is less than or equal to the preset peak value, the periodic fluctuation of the power consumption is relatively weak, and the power consumption characteristics of different industries are relatively independent, at this time, the prediction model is separated to make more accurate prediction for different industries. According to the actual periodic fluctuation of the power consumption, the processing mode is flexibly adjusted, so that the model can maintain good performance in different cases, enhance the adaptability of the model, and further improve the prediction efficiency of the power consumption.

[0055] Specifically, the process of increasing the model training set includes: The data amount of the training set used to train the first prediction model is adjusted to the corresponding value based on the maximum peak value, wherein, Obtain each peak value in the spectrum diagram to obtain the maximum peak value; The increase amplitude of the data amount is negatively correlated with the maximum peak value; The process of separating the prediction model includes: The predicted power consumption output by the model trained by the industrial power consumption data set is determined as the industrial predicted power consumption; The predicted power consumption output by the model trained by the service power consumption data set is determined as the service predicted power consumption; The sum of the industrial predicted power consumption and the service predicted power consumption is determined as the power consumption predicted in the next preset monitoring period.

[0056] Specifically, based on the periodic quantization value of the main analysis data, the model training set is increased or the prediction model is separated, and according to the periodic characteristics of the main industrial power consumption data, two different strategies are adopted. The maximum peak value represents the intensity of the most important periodic component in the power consumption change. When the maximum peak value is small, it means that the periodic fluctuation of the power consumption is weak, and the model needs more data to learn this complex change mode. Therefore, the increase amplitude of the data amount is negatively correlated with the maximum peak value, that is, the greater the maximum peak value, the relatively less the increased data amount, because the model can already capture the main change characteristics from the existing data at this time; the smaller the maximum peak value, the relatively more the increased data amount, to help the model learn more detailed features. By adjusting the training set data amount according to the maximum peak value, the learning effect of the model is guaranteed, and the problems of long training time and overfitting caused by too large data amount are avoided, and the prediction efficiency for the power consumption is further improved.

[0057] In the embodiment, optionally, The peak value is compared with the first preset peak value and the second preset peak value; If the peak value is less than or equal to the first preset peak value, the data amount of the training set used to train the first prediction model is adjusted to 1.31 times of the current data amount; If the peak value is less than or equal to the second preset peak value and greater than the first preset peak value, the data amount of the training set used to train the first prediction model is adjusted to 1.21 times of the current data amount; If the peak value is greater than the second preset peak value, the data amount of the training set used to train the first prediction model is adjusted to 1.11 times of the current data amount; The first preset peak value is 3.2J0, and the second preset peak value is 3.8J0, where J0 is a spectral average value.

[0058] Specifically, the electricity consumption characteristics of the industrial and service industries are different, and the changes in electricity consumption are affected by different factors. When the periodic fluctuations of electricity consumption are relatively weak, training models for industrial and service electricity consumption data sets respectively can more accurately capture the electricity consumption rules of different industries. Finally, the prediction results of the two models are added to obtain the predicted electricity consumption in the next preset monitoring period. The electricity consumption characteristics of different industries are fully considered to avoid the mutual interference of different industrial electricity consumption characteristics and improve the prediction accuracy. By separating the prediction model to model and predict the electricity consumption characteristics of the industrial and service industries respectively, the electricity consumption rules of different industries are more accurately captured, more accurate basis is provided for power supply and management, and the prediction efficiency of electricity consumption is further improved.

[0059] So far, the technical solutions of the present application have been described in combination with the preferred embodiments shown in the drawings, but those skilled in the art can easily understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to related technical features without departing from the principles of the present application, and the technical solutions after these changes or replacements will fall within the protection scope of the present application.

[0060] The above description is only the preferred embodiments of the present application and is not intended to limit the present application; for those skilled in the art, the present application can have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for intelligent prediction of regional electricity consumption based on multi-source heterogeneous data, characterized in that, include: Collect electricity consumption data for each historical preset monitoring period within the region to construct an electricity consumption dataset; Determine the characteristic value of power change based on historical power elasticity coefficient; Based on the power change characterization values, determine the predicted impact of electricity consumption on the next preset monitoring period; When it is determined that the electricity consumption forecast has a strong influence on the next preset monitoring period, the electricity consumption is corrected by the elastic influence weighting coefficient, and the qualification of the electricity consumption forecast is determined based on the energy change characterization value. When identifying anomalies in electricity consumption forecasts, the forecast parameters for electricity consumption in the next preset monitoring cycle are determined based on the quantitative value of regional industrial changes. This includes identifying the main analysis dataset and determining whether to increase the model training set or separate the forecast model based on the periodic quantitative value of the main analysis data. If the next preset monitoring period is determined to be characterized by a weak impact of electricity consumption forecast, the electricity consumption within the next preset monitoring period is determined based on the electricity consumption dataset.

2. The intelligent prediction method for regional electricity consumption based on multi-source heterogeneous data according to claim 1, characterized in that, The process of determining the characteristic value of power change based on historical power elasticity coefficients includes: Plot the time-domain curves of the electroelasticity coefficients based on historical data; The absolute value of the ratio of the absolute value of the slope of the time-domain curve of the power elasticity coefficient at the current time node to the preset stable slope is obtained to obtain the power change characterization value.

3. The intelligent prediction method for regional electricity consumption based on multi-source heterogeneous data according to claim 2, characterized in that, The process of determining the predicted impact of electricity consumption on the next preset monitoring period based on electricity change characterization values ​​includes: When the power change characterization value is less than or equal to the preset power change characterization value, the next preset monitoring period is determined to be a weak influence tendency of power consumption prediction, and the power consumption in the next preset monitoring period is determined based on the power consumption dataset. When the power change indicator value is greater than the preset power change indicator value, if the next preset monitoring period is determined to be when the power consumption forecast has a strong influence tendency, the power consumption will be corrected by the elastic influence weighting coefficient.

4. The intelligent prediction method for regional electricity consumption based on multi-source heterogeneous data according to claim 3, characterized in that, The process of correcting electricity consumption by influencing the weighting coefficient through elasticity includes: The selection range of the elasticity coefficient is determined based on the current electric elasticity coefficient; Each preset monitoring period in the historical data where the electric elasticity coefficient is within the selected range of the elasticity coefficient is determined as the reference period; Obtain the impact ratio for each reference period; The elastic force influence weight coefficient is determined based on each influence ratio.

5. The intelligent prediction method for regional electricity consumption based on multi-source heterogeneous data according to claim 4, characterized in that, The process of determining the values ​​representing energy changes includes: The ratio of natural gas consumption in the most recent preset monitoring period to the average natural gas consumption in each preset monitoring period in historical data is used to obtain the natural gas change rate. The ratio of coal consumption in the most recent preset monitoring period to the average coal consumption in each preset monitoring period in historical data is used to obtain the coal change rate. The ratio of oil consumption in the most recent preset monitoring period to the average oil consumption in each preset monitoring period in historical data is used to obtain the oil change rate. The ratio of electricity consumption in the most recent preset monitoring period to the average electricity consumption in each preset monitoring period in historical data is obtained to obtain the electricity consumption change rate. Calculate the average of the natural gas change rate, coal change rate, and oil change rate to obtain the change in alternative energy sources; The energy change characterization value is obtained by solving the ratio of the absolute value of the difference between the change in alternative energy and the rate of change in electricity consumption to the rate of change in electricity consumption.

6. The intelligent prediction method for regional electricity consumption based on multi-source heterogeneous data according to claim 5, characterized in that, Determining the adequacy of electricity consumption forecasts based on energy change characterization values ​​includes: When the energy change characterization value is less than or equal to the preset energy change characterization value, the prediction of electricity consumption is determined to be qualified, and the product of the elasticity influence weighting coefficient and the electricity consumption predicted by the first prediction model is determined as the predicted electricity consumption for the next preset monitoring period. When the energy change characterization value is greater than the preset energy change characterization value, an anomaly is identified in the prediction of electricity consumption. The prediction parameters for electricity consumption in the next preset monitoring cycle are then determined based on the quantitative value of regional industrial change.

7. The intelligent prediction method for regional electricity consumption based on multi-source heterogeneous data according to claim 6, characterized in that, The process of determining the quantitative value of regional industrial changes includes: Obtain industrial electricity consumption and service electricity consumption within each historical preset monitoring period in the region, so as to obtain industrial electricity consumption datasets and service electricity consumption datasets respectively; The ratio of industrial electricity consumption in the most recent preset monitoring period to the average industrial electricity consumption in each historical preset monitoring period is used to obtain the quantitative value of regional industrial changes. The ratio of the service electricity consumption in the most recent preset monitoring period to the average service electricity consumption in each historical preset monitoring period is used to obtain the quantitative value of changes in the regional service industry.

8. The intelligent prediction method for regional electricity consumption based on multi-source heterogeneous data according to claim 7, characterized in that, The process of determining the predicted parameters for electricity consumption in the next preset monitoring period based on the quantitative value of regional industrial changes includes: When the quantitative value of regional industrial sector change is less than or equal to the quantitative value of regional service sector change, the service electricity consumption dataset will be identified as the primary dataset for analysis. When the quantitative value of regional industrial sector change is greater than the quantitative value of regional service sector change, the industrial electricity consumption dataset will be selected as the primary dataset for analysis.

9. The intelligent prediction method for regional electricity consumption based on multi-source heterogeneous data according to claim 8, characterized in that, The process of determining whether to increase the model training set or separate the prediction model based on the periodic quantization values ​​of the main analysis data includes: Based on the electricity consumption of each preset monitoring period in the main analysis dataset, the main electricity consumption time-domain curves are plotted. The signal of the main power consumption time domain curve is converted into a frequency domain signal by Fourier transform to obtain the spectrum. Obtain the peak values ​​in the spectrum; If a peak value exceeds the preset peak value, the model training set will be increased. When all peak values ​​are less than or equal to the preset peak value, the prediction model is separated.

10. The intelligent prediction method for regional electricity consumption based on multi-source heterogeneous data according to claim 9, characterized in that, The process of increasing the model training set includes: Based on the maximum peak value, the amount of data in the training set used to train the first prediction model is adjusted to a corresponding value, wherein, Obtain each peak in the spectrum to find the maximum peak. The increase in data volume is negatively correlated with the maximum peak value. The process of separating the prediction model includes: The predicted electricity consumption output by the model trained using the industrial electricity consumption dataset is determined as the predicted industrial electricity consumption. The predicted electricity consumption output by the model trained using the service electricity consumption dataset is determined as the service predicted electricity consumption. The sum of the industrial forecasted electricity consumption and the service forecasted electricity consumption is determined as the electricity consumption forecast for the next preset monitoring cycle.

Images