Charging pile load prediction method, device and equipment and computer storage medium

Abstract

The invention discloses a charging pile load prediction method, device and equipment and a computer storage medium, and the method comprises the steps: obtaining the historical load data of a charging pile, and carrying out the normalization processing of the historical load data; training a prediction model by using the normalized historical load data and the influence factors; judging whether an unconventional event exists in the to-be-predicted day or not; if the to-be-predicted day has the unconventional event, matching a similar day corresponding to the to-be-predicted day, and predicting the load value of the charging pile according to the load of the similar day; and if the unconventional event does not exist in the to-be-predicted day, predicting the load value of the charging pile by adopting the trained prediction model. Through the method, the overall load condition of the charging pile can be reflected more accurately.

Description

Technical Field

[0001] The present application relates to the technical field of power systems, and in particular, to a method, device, equipment, and computer storage medium for predicting the load of charging piles. Background Art

[0002] With the continuous growth of the penetration rate of electric vehicles, the strong randomness of their charging loads poses significant challenges to the stable operation of power systems, especially distribution networks. Unordered charging behaviors not only bring many uncertain threats to grid planning, such as the load acceptance capacity of distribution networks and grid security risks, but also may cause severe fluctuations in the power supply quality of distribution networks during large-scale centralized charging, and even result in a situation where the load demand cannot be met in a short period. Therefore, accurately predicting the maximum load power of charging piles is of crucial significance for grid planning, stable operation, orderly charging of electric vehicles, and reasonable planning and construction of charging piles.

[0003] As a special type of time-series data, the load of charging piles not only exhibits non-linear characteristics, but its load situation is also profoundly affected by the current and past performance of charging piles. Currently, the methods for predicting the load of charging piles can be mainly divided into two categories: those based on statistical learning and those based on machine learning. The load prediction method based on statistical learning models and analyzes the historical data of charging piles by using statistical analysis techniques to achieve the prediction of future charging loads. Although this method is easy to implement, its prediction accuracy is often limited by data quality and sample size, making it difficult to ensure high-precision prediction results. The load prediction method based on machine learning mainly focuses on the load prediction of individual charging piles. However, in practical applications, the drastic fluctuations in the load of individual charging piles usually do not have a significant impact on the overall load of the substation area where they are located.

[0004] Therefore, it is urgent to explore and propose an improved method for predicting the load of charging piles to accurately reflect the overall load situation of charging piles in residential communities, thereby providing stronger support for grid planning and stable operation. Summary of the Invention

[0005] In order to overcome the deficiencies of the prior art, the purpose of the present application is to provide a method, device, equipment, and computer storage medium for predicting the load of charging piles, so as to accurately reflect the overall load situation of charging piles.

[0006] The purpose of the present application is achieved by adopting the following technical solutions: In a first aspect, the present application provides a method for predicting the load of a charging pile, including: obtaining the historical load data of the charging pile and performing normalization processing on the historical load data, where the number of charging piles is multiple and they are all located within a predetermined range, and the historical load data includes at least one of charging time and charging amount; obtaining the influencing factors related to the load of the charging pile, where the influencing factors include at least one of weather, temperature, oil price, and date type; training a prediction model using the normalized historical load data and the influencing factors; determining whether there is an unconventional event on the day to be predicted, where the unconventional event is one of weather, temperature, oil price, and date type that meets a preset condition; if there is an unconventional event on the day to be predicted, then matching a similar day corresponding to the day to be predicted and predicting the load value of the charging pile according to the load of the similar day; if there is no unconventional event on the day to be predicted, then using the trained prediction model to predict the load value of the charging pile.

[0007] In one possible implementation manner, the predicting the load value of the charging pile according to the load of the similar day includes: determining whether the date distance between the day to be predicted and the similar day is greater than a preset value; if so, then performing an operation on the load of the similar day using the ratio method to predict the load value of the charging pile.

[0008] In one possible implementation manner, the using the trained prediction model to predict the load value of the charging pile includes: obtaining a preliminary load prediction value of the charging pile using the trained prediction model; determining whether the error index of the preliminary load prediction value meets a preset standard, where the error index includes percentage error and root mean square error; if so, then using the preliminary load prediction value as the prediction result of the load value of the charging pile.

[0009] In one possible implementation manner, the prediction model adopts an LSTM model.

[0010] In one possible implementation manner, the obtaining the historical load data of the charging pile includes: repairing missing values and outliers using the average interpolation method.

[0011] In one possible implementation manner, the obtaining the influencing factors related to the load of the charging pile includes: calculating the correlation between different meteorological factors and the load of the charging pile respectively through the Person correlation coefficient to select the meteorological factors related to the load of the charging pile.

[0012] In one possible implementation manner, the obtaining the influencing factors related to the load of the charging pile further includes: calculating the correlation between different external factors and the load of the charging pile respectively through the grey correlation coefficient to select the external factors related to the load of the charging pile.

[0013] In a second aspect, the present application further provides a charging pile load prediction device, including: a historical data acquisition module, configured to acquire the historical load data of the charging piles and perform normalization processing on the historical load data, where the number of the charging piles is multiple and all are located within a predetermined range, and the historical load data includes at least one of charging time and charging amount; an influencing factor acquisition module, configured to acquire the influencing factors related to the charging pile load, where the influencing factors include at least one of weather, temperature, oil price, and date type; a normalization module, configured to train a prediction model using the normalized historical load data and the influencing factors; an unconventional judgment module, configured to judge whether there is an unconventional event on the day to be predicted, where the unconventional event is one of weather, temperature, oil price, and date type that meets a preset condition; a similar day prediction module, configured to, if there is an unconventional event on the day to be predicted, match the similar day corresponding to the day to be predicted and predict the load value of the charging pile according to the load of the similar day; a model prediction module, configured to, if there is no unconventional event on the day to be predicted, use the trained prediction model to predict the load value of the charging pile.

[0014] In a third aspect, the present application further provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where when the processor executes the computer program, the method described in any one of the first aspects above is implemented.

[0015] In a fourth aspect, the present application further provides a computer-readable storage medium, on which a computer program is stored, where when the computer program is executed by a processor, the method described in any one of the first aspects above is implemented.

[0016] Compared with the prior art, the beneficial effects of the present application at least include: In the charging pile load prediction method of the present application, by collecting and normalizing the historical load data of multiple charging piles, combining various influencing factors such as weather, temperature, oil price, and date type, a prediction model is trained, and for the possible unconventional events on the day to be predicted, the prediction strategy is flexibly switched, and the similar day data is matched for prediction, so as to improve the accuracy of load prediction and provide strong decision-making support for the operation and management of the charging piles.

[0017] In the charging pile load prediction method of the present application, when predicting the charging pile load using the similar day, by judging whether the date distance between the day to be predicted and the similar day exceeds a preset range, when the date distance is too large, the ratio method is used to calculate the load data of the similar day, which can more accurately predict the load value of the charging pile, thereby improving the accuracy of the prediction result.

[0018] In the charging pile load prediction method of the present application, a preliminary load prediction value of the charging pile is obtained through a trained prediction model, and the error index of the prediction result is further evaluated. Only when the error meets the preset standard, the preliminary prediction value is determined as the final load prediction result, which can further improve the accuracy of the prediction result and provide more accurate data support for the load management of the charging pile. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings involved in the embodiments of the present application will be briefly introduced below.

[0020] Figure 1 is a schematic flowchart of a charging pile load prediction method provided by an embodiment of the present application; Figure 2 is a schematic flowchart of a step of a charging pile load prediction method provided by an embodiment of the present application; Figure 3 is a schematic flowchart of another step of a charging pile load prediction method provided by an embodiment of the present application; Figure 4 is a schematic flowchart of yet another step of a charging pile load prediction method provided by an embodiment of the present application; Figure 5 is a schematic flowchart of a charging pile load prediction method provided by another embodiment of the present application; Figure 6 is a structural framework diagram of a charging pile load prediction device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] To facilitate the understanding of the present application, the present application will be described more comprehensively below with reference to the relevant drawings. The preferred embodiments of the present application are shown in the drawings. However, the present application can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, these embodiments are provided to make the understanding of the disclosure of the present application more thorough and comprehensive.

[0022] Embodiment 1 Figure 1 is a schematic flowchart of a charging pile load prediction method provided by an embodiment of the present application. Refer to Figure 1 As shown, the method may include: Step S100, obtain the historical load data of the charging piles and perform normalization processing on the historical load data. Among them, the number of the charging piles is multiple and they are all located within a predetermined range. The historical load data includes at least one of charging time and charging amount. The predetermined range can be determined according to the division of residential areas. For example, the predetermined range can be a residential community, a residential neighborhood, a residential street, etc. In other words, the historical load data of multiple charging piles can be extracted from the charging pile management system or the relevant database to ensure that the data covers all the charging piles located within the predetermined range (such as residential communities, neighborhoods, streets). The data content can include charging time and charging amount, or other relevant information such as charging power and charging times. The average interpolation method can be used to repair missing values and outliers to ensure the integrity and accuracy of the data. Finally, perform normalization processing on the historical load data to convert it to a unified scale for subsequent model training and comparison. The calculation formula for normalization processing can be: X′ = X-X min X max -X min In the formula, X′ represents the data after normalization processing; X represents the data before normalization processing; X max , X min respectively represent the maximum value and the minimum value in the data sequence.

[0023] Optionally, the obtaining of the historical load data of the charging piles may include: using the average interpolation method to repair missing values and outliers. It should be noted that when obtaining the historical load data of the charging piles, the integrity and accuracy of the data are crucial. However, due to various reasons (such as equipment failures, communication interruptions, etc.), there may be missing values and outliers in the data. To improve the data quality, the average interpolation method can be used to repair these problematic data.

[0024] The average interpolation method may specifically include steps such as identifying missing values and outliers, calculating the average value, interpolation repair, verifying the repair effect, and data smoothing processing.

[0025] Among them, missing values mean that in the dataset, the load data at some time points may be empty or not recorded. Outliers mean that the load data at some time points in the dataset significantly deviates from the normal range, which may be caused by equipment failures or data recording errors.

[0026] Calculating the average value means that for the time point where the missing value or outlier is located, select the valid data within a period of time before and after it (for example, one hour before and after) for average calculation. The selection of this time period can be adjusted according to the actual situation to ensure that the selected data is representative. The specific calculation formula for calculating the average value can be: x i =(x i-1 +x i+1 ) / 2 wherein, x i is the data after repair or completion, and x i-1 and x i+1 are the data before and after the null value or abnormal data respectively.

[0027] Interpolation repair, that is, using the calculated average value as the estimated value of the missing value or abnormal value to replace the missing value or abnormal value in the original data.

[0028] Verify the repair effect, that is, after interpolation repair, verify the repaired data to ensure that the repaired data is logically reasonable and does not introduce new abnormal values.

[0029] Data smoothing processing, that is, if the repaired data still appears abrupt in terms of changes, a smoothing processing algorithm such as the moving average method can be considered to further smooth the data.

[0030] Repairing missing values and abnormal values by the average interpolation method can maintain the continuity of the data and avoid data breaks caused by missing values or abnormal values. The repaired data is more accurate and reliable, which is beneficial to subsequent data analysis and processing.

[0031] Step S200, obtain the influencing factors related to the charging pile load, where the influencing factors include at least one of weather, temperature, oil price, and date type. In other words, factors such as weather (such as sunny, rainy, snowy days, etc.), temperature, oil price, and date type (such as weekdays, weekends, holidays) can be considered for their influence on the charging pile load. The linear correlation between different meteorological factors and the charging pile load can be evaluated by calculating the Pearson correlation coefficient, and the meteorological factors with higher correlation can be selected as the model input. The non - linear correlation between different external factors (such as oil price, date type, etc.) and the charging pile load can be evaluated by calculating the grey correlation coefficient, and the external factors with higher correlation can be selected as the model input.

[0032] Figure 2 is a schematic flow chart of step S200 of the charging pile load prediction method provided by an embodiment of the present application. Refer to Figure 2 As shown, the obtaining of the influencing factors related to the charging pile load may specifically include: Step S210, obtain the correlation between different meteorological factors and the charging pile load respectively by calculating the Person correlation coefficient, so as to select the meteorological factors related to the charging pile load. Among them, the Person correlation coefficient, that is, the Pearson Correlation Coefficient, is also called the Pearson correlation coefficient.

[0033] Specifically, the Person correlation coefficient formula can be used to calculate the correlation coefficient between each meteorological factor and the charging pile load. The value range of the Person correlation coefficient is between -1 and 1. The closer its absolute value is to 1, the stronger the linear relationship between the two variables; the closer it is to 0, the weaker the linear relationship. According to the calculated correlation coefficient, a threshold value (such as 0.5 or higher) can be set, and the meteorological factors with correlation coefficients exceeding the threshold value can be selected as factors significantly correlated with the charging pile load. The positive and negative signs of the correlation coefficient can also be considered to determine the positive or negative correlation between the meteorological factors and the load. The Person correlation coefficient formula can be: Among them, r represents the Person correlation coefficient between variables x and y, reflecting the degree of correlation between the linear relationship between load influencing factors; x and y are two variables whose correlation coefficients are to be calculated.

[0034] Through step S210, meteorological factors closely related to the charging pile load can be identified, providing a strong basis for subsequent load forecasting, charging pile optimization layout and operation strategy formulation.

[0035] See also Figure 2 As shown, the acquisition of influencing factors related to the charging pile load may further include: Step S220, by calculating the grey correlation coefficient, the correlation between different external factors and the charging pile load is obtained to select the external factors related to the charging pile load. That is, by calculating the grey correlation coefficient, the correlation between different external factors and the charging pile load is evaluated respectively, which may specifically include the steps of determining the analysis series, dimensionless processing of data, calculating the grey correlation coefficient, calculating the grey correlation degree, correlation degree sorting and factor selection.

[0036] Among them, determine the analysis series. The analysis series includes determining the reference series (parent series) and determining the comparison series (subsequence). The reference series is a data series that reflects the characteristics of the charging pile load. The historical load data of the charging pile is usually selected as the reference series. The comparison series is a data series composed of external factors that affect the charging pile load. These external factors may include meteorological factors such as temperature, humidity, wind speed, precipitation, and non-meteorological factors such as charging pile layout, site availability, power supply capacity, and electric vehicle ownership.

[0037] Dimensionless processing of data. Due to the different physical meanings of various factors in the system, the dimensions of the data are not necessarily the same, which makes it inconvenient to compare or difficult to get the correct conclusion when comparing. Therefore, before conducting grey relational analysis, dimensionless processing of the data can be performed. Dimensionless processing can be performed as initialization processing or averaging processing.

[0038] Calculate the grey correlation coefficient. The grey correlation coefficient is used to measure the correlation degree between the comparison sequence and the reference sequence at each moment (or each data point). The calculation formula for the grey correlation coefficient can be: Δ = |y(k) - x i (k)| where ξ i (k) is the correlation coefficient; Δ is the absolute difference between each point on each comparison sequence x i and each point on the parameter sequence y; Δ min is the minimum value of all absolute differences; Δ max is the maximum value of all absolute differences; ρ is the resolution coefficient, taking 0.5.

[0039] Calculate the grey correlation degree. Since the correlation coefficient is the correlation degree value between the comparison sequence and the reference sequence at each moment, the information is too scattered to facilitate overall comparison. Therefore, the correlation coefficients at each moment can be concentrated into one value, that is, find their average value, as a quantitative representation of the correlation degree between the comparison sequence and the reference sequence, that is, the correlation degree. The larger the correlation degree, the stronger the correlation between the external factor and the charging pile load. Calculate the correlation degree, which can be specifically realized through the following formula: where γ o i is the correlation degree between the reference sequence and the i-th comparison sequence; m is the length of the sequence, that is, the number of evaluation indicators.

[0040] Correlation degree sorting and factor selection. The calculated correlation degrees can be sorted by size. The larger the correlation degree of the external factor, the stronger the correlation with the charging pile load. According to actual needs and threshold settings, select the external factors that are significantly correlated with the charging pile load.

[0041] Step S300, use the normalized historical load data and the influencing factors to train the prediction model. Specifically, according to the data characteristics and prediction requirements, a suitable prediction model can be selected, such as linear regression, support vector machine, neural network, etc.

[0042] Preferably, the prediction model can be constructed based on the LSTM model. It should be noted that LSTM, that is, the long short-term memory network, is the abbreviation of Long Short-Term Memory. It is a special recurrent neural network (RNN). By introducing memory cells (Cell State) and a series of gating mechanisms (input gate, forget gate and output gate), it can solve the problem of gradient disappearance in the processing of long sequence data by ordinary RNNs.

[0043] After constructing the prediction model, the normalized historical load data and the selected influencing factors can be used as inputs to train the prediction model to obtain model parameters.

[0044] Step S400: Determine whether there is an unconventional event on the day to be predicted. The unconventional event is one of weather, temperature, oil price, and date type that meet the preset conditions. Specifically, for weather, specific weather conditions can be set as unconventional events, such as extreme weather like heavy rain, blizzard, typhoon, etc.; for temperature, a temperature threshold can be set. For example, when the temperature is higher or lower than a specific value, it is regarded as an unconventional event; for oil price, a threshold for oil price change can be set. For example, when the oil price rises or falls by more than a certain percentage, it is regarded as an unconventional event; for date type, a specific list of date types can be formulated, and holidays, large event days, etc. are listed as unconventional events.

[0045] Specifically, first obtain the relevant information of the day to be predicted. Among them, weather information and temperature information can be obtained through weather forecast APIs or meteorological data services; oil price information can be obtained from oil price data services or relevant institutions; date type information can be obtained by checking whether the day to be predicted is a specific date type such as a holiday or a large event day.

[0046] Then, determine whether there is an unconventional event. Specifically, the weather condition of the day to be predicted can be compared with the unconventional weather conditions to determine whether the weather meets the conditions of the unconventional event; the predicted temperature value of the day to be predicted can be compared with the set temperature threshold to determine whether it exceeds the range to determine whether the temperature meets the conditions of the unconventional event; the oil price change situation can be calculated and compared with the set oil price change threshold to determine whether the oil price meets the conditions of the unconventional event; it can be checked whether the day to be predicted is one of the defined specific date types to determine whether the date type meets the conditions of the unconventional event.

[0047] Step S500: If there is an unconventional event on the day to be predicted, match the similar day corresponding to the day to be predicted, and predict the load value of the charging pile according to the load of the similar day. In other words, the similarity of factors such as weather, temperature, and date type can be considered to find a date similar to the day to be predicted in the historical data. The effectiveness of the similarity can be evaluated by determining whether the date distance between the day to be predicted and the similar day is greater than the preset value. And when the date distance is large, the ratio method is used to calculate the load of the similar day to adapt to the special situation of the day to be predicted and predict the load value of the charging pile.

[0048] Figure 3 It is a schematic flowchart of step S500 of the charging pile load prediction method provided by an embodiment of the present application. See Figure 3As shown, predicting the load value of a charging pile based on the load on similar days may specifically include: Step S510: Determine whether the date distance between the day to be predicted and the similar day is greater than a preset value. The date distance between the day to be predicted and each similar day can be calculated by counting the number of days, and then the calculated date distance is compared with the preset threshold. This preset value can be determined based on experience or data analysis to judge whether the date distance is close enough. Only by way of example, the preset value can be 3 days, 5 days, 10 days, 20 days, 30 days, etc.

[0049] Step S520: If so, use the ratio method to operate on the load of the similar day to predict the load value of the charging pile. Specifically, one or more similar days relatively close to the day to be predicted can be selected as the reference days. According to the load data of the reference days and the changes in relevant factors (such as weather, temperature, oil price, etc.) between the day to be predicted and the reference days, calculate the ratio coefficient, which reflects the possible change ratio of the load from the reference day to the day to be predicted. Finally, multiply the load data of the reference day by the ratio coefficient to obtain the load prediction value of the day to be predicted, which can be specifically realized through the following calculation formula: where p i and p j are the maximum loads of the day to be predicted and the similar day respectively, is the load growth coefficient, and are the average values of the maximum loads in the previous K days of the day to be predicted and the similar day respectively, and K can be selected according to the actual situation.

[0050] Step S600: If there are no abnormal events on the day to be predicted, use the trained prediction model to predict the load value of the charging pile. In other words, a trained preset model can be used to input the normalized influencing factor data of the day to be predicted to obtain the preliminary load prediction value of the charging pile, and then calculate the error indicators of the preliminary load prediction value, such as percentage error and root mean square error, to evaluate the accuracy of the prediction result. If the error indicator meets the preset standard, the preliminary load prediction value is used as the prediction result of the load value of the charging pile; if not, the model can be adjusted or retrained.

[0051] In the charging pile load prediction method provided in this embodiment, by collecting and normalizing the historical load data of multiple charging piles, combining various influencing factors such as weather, temperature, oil price, and date type, a prediction model is trained, and for possible abnormal events on the day to be predicted, the prediction strategy is flexibly switched to match the similar day data for prediction, thereby improving the accuracy of load prediction and providing strong decision-making support for the operation and management of charging piles.

[0052] Figure 4It is a schematic flowchart of step S600 of the charging pile load prediction method provided by an embodiment of the present application. Refer to Figure 4 As shown, using the trained prediction model to predict the load value of the charging pile may specifically include: Step S610, obtaining a preliminary load prediction value of the charging pile using the trained prediction model. Specifically, relevant data of the day to be predicted (such as influencing factors like date type, weather, temperature, oil price, etc.) is input into the prediction model. The prediction model is used for calculation to obtain the preliminary load prediction value of the charging pile. This preliminary load prediction value is obtained based on the learning of historical data by the prediction model and the analysis of the current input data.

[0053] Taking the prediction model based on the LSTM model as an example, the LSTM layer in the LSTM model (a layer structure composed of multiple LSTM units) can be used to learn the characteristics of multivariate historical data, so as to predict multivariate loads. Then, the inverse normalization process is performed on the prediction result as the preliminary load prediction value of the prediction model. Inverse normalization is the inverse operation of normalization, and its calculation formula can be: X = X' * (X max - X min ) + X min where X represents the data after inverse normalization; X' represents the data before inverse normalization; X max , X min represent the maximum and minimum values in the data sequence respectively.

[0054] Step S620, determining whether the error index of the preliminary load prediction value meets the preset standard. The error index includes percentage error and root mean square error. In other words, calculate the error index between the preliminary load prediction value and the target load value. The target load value can be the actual load value. It should be noted that since the day to be predicted is a future date, when the day to be predicted has not arrived, the actual load value can be replaced by the load value of a similar day in step S500. The error index can include percentage error and root mean square error. Among them, the percentage error, that is, MAPE, full name Mean Absolute Percentage Error, also known as the average absolute percentage error, reflects the relative deviation between the predicted value and the actual value; the root mean square error, that is, RMSE, full name Root Mean Square Error, reflects the square root of the average of the sum of squares of the differences between the predicted value and the actual value, and can measure the overall deviation between the predicted value and the actual value. The preset standard is the preset error standard, and the error standard is determined according to factors such as the performance of historical data and the prediction model. The calculation formulas for percentage error and root mean square error can be respectively: Among them, E MAPE and E RMSE are the absolute percentage error and the root mean square error respectively; y i is the actual value, is the predicted value; n is the number of samples.

[0055] Step S630, if so, use the preliminary load prediction value as the load value prediction result of the charging pile. Specifically, if the error index of the preliminary load prediction value meets the preset standard, that is, the error is within an acceptable range, then it can be considered that the prediction result is reliable. Therefore, the preliminary load prediction value can be used as the load value prediction result of the charging pile.

[0056] It should be noted that even if the prediction result meets the standard, the performance of the prediction model can be continuously monitored and necessary adjustments and improvements can be made according to the actual feedback. If the prediction result does not meet the standard, the model can be retrained, the model parameters can be adjusted, or new influencing factors can be considered.

[0057] Embodiment 2 Figure 5 is a schematic flowchart of a charging pile load prediction method provided by another embodiment of the present application.

[0058] See Figure 5 As shown, before implementing the charging pile load prediction method, the prediction purpose and content can be clarified, that is, the purpose and content to be achieved by the charging pile load prediction method are clarified. For example, clarify that the load prediction is for optimizing the operation of the charging station, improving the utilization rate of charging facilities, preventing overload situations, or for supporting grid dispatching, etc., and clarify the time range of the prediction (such as short-term prediction of the load within a few hours, medium-term prediction of the load within a few days, long-term prediction of the load trend for several months or years), the accuracy requirements of the prediction, and whether it is necessary to predict the load of different types of charging piles (such as fast charging, slow charging), etc.

[0059] See Figure 5 As shown, the charging pile load prediction method provided by the embodiment of the present application can be specifically implemented according to the following steps: Charging pile load data collection and processing. That is, collect the historical load data of charging piles and perform preprocessing to ensure the accuracy and integrity of the data. Specifically, data can be collected for residential charging piles in residential communities, including the time of the maximum charging load per day, the value of the maximum charging load, the daily maximum and minimum temperatures, precipitation, climate and other data. Due to the influence of various random factors, the collected load and load influencing factor data may have missing or abnormal values (such as 0 or null values). If data with missing or abnormal values is used, the prediction results will be inaccurate to a certain extent and the accuracy will be low. Therefore, before performing load prediction, it is necessary to preprocess and correct the data, eliminate data with too short charging times, and use the average interpolation method to repair and fill in null values, 0 values or abnormal data. The calculation formula of the average interpolation method is: x i =(x i-1 +x i+1 ) / 2 In the formula, x i is the data after repair or filling; x i-1 and x i+1 are the data before and after the null value or abnormal data respectively.

[0060] Data preprocessing, selection of load influencing factors. That is, on the basis of data preprocessing, select the factors that affect the load, including meteorological factors and external factors selected through grey relational analysis. Among them, the meteorological factors can be selected using the person coefficient, and the external factors can be selected using the grey relational degree coefficient.

[0061] Specifically, in power load prediction, the influence of influencing factors is considered. By adding appropriate influencing factors to the prediction, the prediction accuracy can be improved. External factors such as temperature, weather, oil price, and date type can be considered in the load prediction of charging piles.

[0062] Among them, for non-linear influencing factors such as date type and daily oil price, calculate the grey relational degree coefficient to select the factors that have a greater impact on the maximum load, and set appropriate coefficient thresholds to filter out irrelevant features.

[0063] First, calculate the correlation coefficient between y(k) and x i (k), and the calculation formula is as follows: Δ = |y(k) - x i (k) In the formula, ξ i (k) is the correlation coefficient; Δ is the absolute difference between each point on each comparison sequence x i and each point on the parameter sequence y; Δ min is the minimum value of all absolute differences; Δ maxis the maximum value of all absolute differences; ρ is the resolution coefficient, taking 0.5.

[0064] Then, calculate the grey correlation degree. It should be noted that the correlation coefficient is the value of the correlation degree between the comparison sequence and the reference sequence at each moment (i.e., each point on the curve), and there may be more than one. However, the information is too scattered to facilitate overall comparison. Therefore, the correlation coefficients at each moment (i.e., each point on the curve) can be concentrated into one value, that is, taking the average value, as a quantitative representation of the correlation degree between the comparison sequence and the reference sequence, as shown in the following formula: In the formula, γ o i is the correlation degree between the reference sequence and the i-th comparison sequence; m is the length of the sequence, that is, the number of evaluation indicators.

[0065] For meteorological factors such as temperature, wind speed, rainfall, rain and snow, the Person method can be used to extract the key meteorological factors that affect the charging pile load. The Pearson correlation coefficient (Person) is the product of the covariance between two load variables divided by the standard deviation, and its value range is [-1, 1]. Its calculation formula is as follows: In the formula, R represents the Person correlation coefficient between variables x and y, reflecting the degree of linear relationship correlation between load influencing factors; x and y are two variables for which the correlation coefficient is to be calculated.

[0066] Construct a maximum load prediction model. That is, based on the charging pile load data and the selected influencing factors, an LSTM model is constructed for the prediction of the maximum load.

[0067] Specifically, use the historical load data of the charging pile, the selected meteorological data, and the date type (season type, week type, holiday) as direct influencing factors to design a prediction model to construct an LSTM model.

[0068] Before predicting the charging pile load data, since the dimensions of the load data are different, preprocessing such as normalizing the load historical data can be performed. Using normalized data can facilitate model training and prediction. The normalization processing formula is as follows: In the formula, X′ represents the data after normalization processing; X represents the data before normalization processing; X max ,X min respectively represent the maximum and minimum values in the data sequence.

[0069] The role of the LSTM layer in the LSTM model is to learn the features of multivariate historical data and predict the multivariate load. The prediction result can be de-normalized as the preliminary prediction result of the load prediction model. De-normalization is the inverse operation of normalization and can be achieved through the following formula: X = X' * (X max - X min ) + X min In the formula, X represents the data after de-normalization; X' represents the data before de-normalization; X max , X min represent the maximum and minimum values in the data sequence respectively.

[0070] Model parameter estimation. That is, estimate the parameters of the LSTM model.

[0071] Perform load prediction. That is, use this model to perform load prediction.

[0072] Whether the error requirement is met. That is, check whether the prediction result meets the error requirement. If not, correct and improve the model, and then re-perform the prediction until the error standard is met.

[0073] Specifically, the absolute percentage error and root mean square error can be used as evaluation indicators, and the error probability density distribution statistical analysis model is used to analyze the prediction performance of the model. The absolute percentage error is used to analyze the influence of extreme abnormal data in the prediction, and the root mean square error describes the error mean from the perspective of percentage. Their calculation formulas are as follows: In the formula, E MAPE and E RMSE are the absolute percentage error and root mean square error respectively; y i is the actual value, is the predicted value; n is the number of samples.

[0074] Finally, output the maximum load prediction result. That is, output the maximum load prediction result that meets the error requirement.

[0075] It should be noted that before the steps of constructing the maximum load prediction model, the impacts of electricity price policies, sudden bad weather, traffic control, etc. on the maximum load of charging piles can be considered, and the combined model is constructed using the similar day correction method to predict the maximum load power of charging piles. That is, it can be judged whether there are special circumstances on the prediction day. If so, the similar day correction prediction is adopted, that is, the prediction is corrected through the data of similar days to improve the prediction accuracy.

[0076] Specifically, for unconventional events, retrieve them from the historical charging pile load database over the years and obtain the prediction results using the method of similar day correction. The basic principle of similar days is that for two days with relatively similar influencing factors such as meteorological characteristics and date types within a certain time span, their load conditions are also relatively similar. Use relevant methods to find the historical similar days of the day to be predicted and correct their loads to obtain the final prediction results.

[0077] When the date distance between the similar day and the prediction day is relatively large, the growth trend of the load can also be considered, such as the influence of factors such as the difference in meteorological conditions in the days before and after the day of the sudden situation. Instead of using the load of the similar day as the prediction result of the day to be predicted, correct it according to the actual situation. The ratio method can be used for correction, which can be specifically implemented through the following formula: In the formula, p i and p j are the maximum loads of the day to be predicted and the similar day respectively, is the load growth coefficient, and are the average values of the maximum loads in the previous K days of the day to be predicted and the similar day respectively, and K can be selected according to the actual situation.

[0078] If it does not exist, the LSTM model is used for load prediction.

[0079] Embodiment 3 Figure 6 is the structural framework diagram of a charging pile load prediction device provided by an embodiment of the present application. Refer to Figure 6 As shown, the device may include: A historical data acquisition module 100, configured to acquire the historical load data of the charging pile and perform normalization processing on the historical load data. Among them, the number of charging piles is multiple and all are located within a predetermined range, and the historical load data includes at least one of charging time and charging amount.

[0080] An influencing factor acquisition module 200, configured to acquire the influencing factors related to the charging pile load, where the influencing factors include at least one of weather, temperature, oil price, and date type.

[0081] A normalization module 300, configured to train a prediction model using the normalized historical load data and the influencing factors; an unconventional judgment module 400, configured to judge whether there is an unconventional event on the day to be predicted, where the unconventional event is one of weather, temperature, oil price, and date type that meets the preset conditions.

[0082] The similar day prediction module 500 is configured to match a similar day corresponding to the day to be predicted and predict the load value of the charging pile according to the load of the similar day if there is an abnormal event on the day to be predicted.

[0083] The model prediction module 600 is configured to predict the load value of the charging pile by using a trained prediction model if there is no abnormal event on the day to be predicted.

[0084] It should be noted here that the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the content disclosed in the method of the above embodiments.

[0085] Embodiment 4 An embodiment of the present application further provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the charging pile load prediction method in the above Embodiment 1 is implemented. Specific examples in this embodiment may refer to the examples described in the above Embodiment 1, and will not be repeated here.

[0086] Embodiment 5 An embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. The computer program, when executed by a processor, implements the charging pile load prediction method in the above Embodiment 1. Similarly, specific examples in this embodiment may refer to the examples described in the above Embodiment 1, and will not be repeated here.

[0087] In addition, those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0088] The above embodiments only represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A charging pile load prediction method, characterized in that: include: Acquire historical load data of a charging pile, and normalize the historical load data, wherein there are multiple charging piles, and all of them are within a predetermined range, and the historical load data includes at least one of a charging time and a charging amount; Obtaining influencing factors related to the charging pile load, wherein the influencing factors include at least one of weather, temperature, oil price, and date type; Using the normalized historical load data and the influencing factors to train a prediction model; Determine whether there is an unconventional event on the day to be predicted, where the unconventional event is one of the weather, temperature, oil price, and date types that meet the preset conditions; If there is an unconventional event on the day to be predicted, a similar day corresponding to the day to be predicted is matched, and the load value of the charging pile is predicted based on the load on the similar day; If there are no unusual events on the predicted day, the trained prediction model is used to predict the load value of the charging pile.

2. The method according to claim 1, characterized in that The method of predicting the load value of the charging pile according to the load on similar days includes: Determine whether the date distance between the predicted day and the similar day is greater than the preset value; If so, the multiple ratio method is used to calculate the load on similar days to predict the load value of the charging pile.

3. The method according to claim 2, characterized in that The method of using the trained prediction model to predict the load value of the charging pile includes: Use the trained prediction model to obtain the initial prediction value of the charging pile load; Determine whether the error index of the preliminary load prediction value meets the preset standard, wherein the error index includes percentage error and root mean square error; If yes, the preliminary load prediction value is used as the load value prediction result of the charging pile.

4. The method according to claim 3, characterized in that The prediction model adopts the LSTM model.

5. The method according to claim 1, characterized in that The obtaining of historical load data of the charging pile includes: The average interpolation method is used to repair missing values ​​and outliers.

6. The method according to any one of claims 1 to 5, characterized in that: The obtaining of influencing factors related to the charging pile load includes: By calculating the Person correlation coefficient, the correlation between different meteorological factors and the charging pile load is obtained to select the meteorological factors related to the charging pile load.

7. The method according to claim 6, characterized in that The obtaining of influencing factors related to the charging pile load also includes: By calculating the grey correlation coefficient, the correlation between different external factors and the charging pile load is obtained to select the external factors related to the charging pile load.

8. A charging pile load prediction device, characterized in that: include: A historical data acquisition module, used to acquire historical load data of a charging pile and normalize the historical load data, wherein the number of the charging piles is multiple and all are within a predetermined range, and the historical load data includes at least one of a charging time and a charging amount; An influencing factor acquisition module, used to acquire influencing factors related to the charging pile load, wherein the influencing factors include at least one of weather, temperature, oil price, and date type; A normalization module, used for training a prediction model using normalized historical load data and the influencing factors; An unconventional judgment module is used to judge whether there is an unconventional event on the predicted day, where the unconventional event is one of the weather, temperature, oil price, and date type that meets the preset conditions; A similar day prediction module is used to match a similar day corresponding to the day to be predicted if there is an unconventional event on the day to be predicted, and predict the load value of the charging pile according to the load on the similar day; The model prediction module is used to use the trained prediction model to predict the load value of the charging pile if there is no unusual event on the predicted day.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

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

Images