An artificial intelligence-based supply chain demand forecasting and dynamic optimization system

Abstract

The application discloses a supply chain demand prediction and dynamic optimization system based on artificial intelligence, and relates to the technical field of charging piles. The perception module is used for collecting the operation data of the charging pile in real time, and specifically includes the daily charging frequency of the single pile and the peak power duration. The perception module is also used for dynamically obtaining the new energy vehicle growth data in the region and obtaining the vehicle battery capacity distribution data. It is also used for periodically receiving the extreme weather warning information released by the meteorological department. According to the long-term demand prediction, the number of new charging piles is determined, and the corresponding number of charging piles is installed in the next long-term period. According to the medium-term demand prediction, the power reserve and demand of each charging pile in the region are monitored in real time, and the distribution of the charging pile is dynamically adjusted. According to the short-term demand prediction, the dynamic time pricing of the charging pile is calculated to guide users to reasonably use the charging pile in different time periods. The utilization rate of the charging pile can be improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of charging piles, in particular to a supply chain demand prediction and dynamic optimization system based on artificial intelligence. BACKGROUND

[0002] It should be noted that, as the possession rate of new energy vehicles rises, most people use new energy vehicles for short-distance travel such as commuting to work. Due to the low endurance characteristics of new energy vehicles, large industrial areas will install charging stations, such as industrial areas and high-tech areas in cities, to provide enough charging piles to meet the charging needs of employees or regional visitors.

[0003] Due to the difference in electricity prices at night and during the day, charging stations will use low valley electricity prices at night to charge and provide regional new energy vehicles with charging piles equipped with energy storage devices to store electricity at night and discharge during the day, thereby reducing regional power consumption costs. However, most employees will charge after arriving at the region during the morning rush hour, so as to avoid the superposition of regional power consumption peaks causing transformer overload. Therefore, it is necessary to predict charging demand in advance and develop subsequent measurements based on the prediction results.

[0004] However, in actual use, the prediction results are short-term predictions, and in seasons with frequent extreme weather, the use demand cannot be met according to the previous prediction. SUMMARY

[0005] To solve the above problems, the application provides a supply chain demand prediction and dynamic optimization system based on artificial intelligence.

[0006] A supply chain demand prediction and dynamic optimization system based on artificial intelligence, characterized by comprising a perception module, the perception module being used for real-time collection of operation data of charging piles, and specifically comprising single-pile daily charging frequency and peak power duration;

[0007] The perception module is also used for dynamic acquisition of regional new energy vehicle growth data and acquisition of vehicle battery capacity distribution data.

[0008] It is also used for periodic reception of extreme weather warning information issued by a meteorological department, and the perception module transmits all data.

[0009] It further comprises a demand prediction module, the demand prediction module being used for reception of data of the perception module and obtaining of a demand prediction result after a prediction period according to the data of the perception module, the demand prediction result specifically comprising decomposition of the prediction period into three time periods of long-term, medium-term and short-term, obtaining of long-term demand prediction, medium-term demand prediction and short-term demand prediction.

[0010] The dynamic optimization module is used for receiving data of the perception module and demand prediction results of the demand prediction module, and performing corresponding dynamic optimization according to long-term period demand prediction, medium-term period demand prediction and short-term period demand prediction in the demand prediction results.

[0011] Preferably, the specific working steps of the perception module are as follows:

[0012] For real-time collection of operation data of the charging pile, the operation data of the charging pile are acquired in real time through sensors and data acquisition equipment installed on the charging pile, including single-pile daily charging frequency Q and peak power duration T.

[0013] The specific acquisition steps include: first, the data acquisition equipment of the charging pile records the start and end time of each charging in real time, and generates a charging record including the start time and end time of charging and other information whenever a vehicle starts charging; at the end of the statistical period, the system will count the charging frequency Q of the charging pile in this day.

[0014] A power sensor is installed in the charging pile to monitor the power output in real time during the charging process. The control system of the charging pile records the power change data over time. When the power reaches the peak value, the timing starts, and stops when the power drops below the peak value, so as to obtain the duration T of the peak power.

[0015] Preferably, the specific working steps of the perception module further include:

[0016] Through data sharing with the vehicle management system, the growth data of new energy vehicles in the dynamic range are obtained, and the vehicle battery capacity distribution data are obtained by using the V2X communication protocol.

[0017] The battery capacity distribution data specifically include average battery capacity V sales , and average battery capacity V sales is obtained by summing up the battery capacities of all registered vehicles in the region and the total number of vehicles.

[0018] For periodic reception of extreme weather warning information issued by the meteorological department, 72-hour extreme weather warning information is obtained through the local meteorological department.

[0019] Preferably, the specific working steps of the demand prediction module are as follows:

[0020] For long-term period demand prediction, firstly, according to the obtained new energy vehicle sales data in the region in the past three long-term periods, the three long-term period new energy vehicle sales data is normalized to the interval [0, 1], 64 LSTM units are selected, a Dropout layer is added to reduce overfitting, a Dense layer is connected, and the next long-term period new energy vehicle sales data is output, the model is completed, and the normalized three long-term period new energy vehicle sales data is input into the completed model, and the next long-term period new energy vehicle sales data P is output new ;

[0021] Then, according to the formula , the basic demand prediction Qh of each long-term period is calculated L , wherein Δ grid is the influence of power grid expansion;

[0022] Then, the basic demand prediction Qh of each long-term period L is subtracted from the actual demand of each long-term period, and the historical residual ∈1 is obtained;

[0023] An MLP neural network model is selected, all obtained historical residuals ∈1 data are divided into a training set, a validation set and a test set, and are divided in proportions of 70%, 15% and 15%, a residual model is trained, historical residuals ∈1 are input into the residual model, and the next long-term period residual ∈ is obtained;

[0024] Then, according to the formula , the long-term period demand prediction Q L is calculated.

[0025] Preferably, the specific working steps of the demand prediction module further include:

[0026] For medium-term period demand prediction, according to the formula

[0027]

[0028] , the medium-term demand Q M is calculated, wherein σ is a regional economic vitality factor, S season is a seasonal fluctuation coefficient, A is a charging station radiation capacity value, μ is a distance attenuation coefficient, and R is the number of charging piles within the range of the charging station;

[0029] , wherein e -μd is a distance attenuation factor, which is used to describe the phenomenon that the charging demand gradually weakens with the increase of distance, and d is a distance attenuation coefficient of the distance from the region where the charging station is located.

[0030] Preferably, the specific working steps of the demand prediction module further include:

[0031] For regional economic vitality factor p, is obtained, where 1 is the annual GDP of the city where the region is located, 2 is the annual total GDP of the country, N1 is the number of newly registered cars in the city where the region is located within a period of time, and N2 is the total number of cars in the city where the region is located;

[0032] For seasonal fluctuation coefficient S season , is obtained based on the fitting of the charging amount data of the same period in the past three years, specifically, is obtained, where t represents the month;

[0033] For charging station radiation capacity value A, according to the formula is obtained, where P t is the average daily passenger flow in the range of the charging station, D p is the charging pile density, and N b is the number of bus lines near the region.

[0034] Preferably, the specific working steps of the demand prediction module further include:

[0035] For short-term period demand prediction, the specific way is:

[0036] Q S = Q b M *(1+0.3ΔT+0.4R r +0.3W S )*(1+0.15*ΔC)*X

[0037] Where ΔT is the difference between the current temperature and the standard temperature, R r is the current rainfall intensity, W S is the current wind speed, ΔC is the battery capacity attenuation rate, which is obtained by subtracting the current battery capacity from the initial battery capacity and then dividing by the initial battery capacity, and X is the traffic congestion value, which is obtained by is obtained, where TL is the traffic index, is the baseline value of charging demand, i.e., the basic level of charging demand without other influencing factors.

[0038] Preferably, the specific working steps of the dynamic optimization module include the following:

[0039] Every other period, the prediction error E of the short-term period demand prediction is calculated, specifically according to the formula: The prediction error E is calculated and obtained, where Q1 is the actual charging demand value, and if E is greater than 15%, then The weight coefficients 0.3, 0.3 and 0.4 in the formula are adjusted by 2% if E is greater than 0, and the weight coefficients 0.3, 0.3 and 0.4 are adjusted by 2% if E is less than 0.

[0040] Preferably, the specific working steps of the dynamic optimization module further include the following:

[0041] For a long-term period, the number of newly added charging piles N is calculated according to the formula In the next long-term period, the number of charging piles N is installed in the region according to the number of newly added charging piles N, wherein P x is the power of a single charging pile, and T1 is the average working time of the charging pile.

[0042] In the medium-term period, the real-time electric energy reserve data E and the safety electric energy threshold E1 of each charging pile in the region are obtained from the perception module, and the short-term period, medium-term period and long-term period demand prediction Q L , Q M and Q S of each charging pile are obtained from the demand prediction module.

[0043] The inventory monitoring degree index H of each region is calculated, and when the inventory monitoring degree index H is greater than a preset threshold, the movable charging pile is moved to the region.

[0044] In the short-term period, the dynamic time pricing N j of the charging pile in the short-term period is calculated according to the formula N L = P L *(1+0.1*Q j ), wherein P L is the basic electricity price.

[0045] Preferably, the specific steps for calculating the inventory monitoring degree index H of the charging pile are as follows:

[0046] The inventory monitoring degree index H of each charging pile is calculated by H=(E-E1)+(Q L *0.2)+(Q M *0.1)+(Q S *0.02).

[0047] Beneficial effects: according to long-term demand prediction, determine the number of new charging piles, and install the corresponding number of charging piles in the next long-term period; according to medium-term demand prediction, real-time monitor the power reserve and demand of each charging pile in the region, dynamically adjust the distribution of charging piles; according to short-term demand prediction, calculate the dynamic time pricing of charging piles, guide users to reasonably use charging piles in different time periods; can improve the utilization rate of charging piles, reduce investment cost, ensure enough charging resources in demand high time and place, improve user experience. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 is a flowchart of the present application. DETAILED DESCRIPTION

[0049] As shown in Figure 1 : a supply chain demand prediction and dynamic optimization system based on artificial intelligence, comprising a perception module, the perception module is used for real-time collection of operation data of charging piles, specifically including single pile daily charging times and peak power duration;

[0050] The perception module is also used for dynamically obtaining new energy vehicle growth data in the region, and obtaining vehicle battery capacity distribution data;

[0051] Also used for periodically receiving extreme weather warning information issued by the meteorological department, and the perception module transmits all data; it should be noted that real-time collection of operation data of charging piles, including single pile daily charging times, peak power duration and fault type code, can accurately master the use frequency, performance and fault of charging piles, provide basic data support for subsequent operation and maintenance and management decision;

[0052] Dynamically obtaining new energy vehicle growth data in the region and vehicle battery capacity distribution data is helpful to understand the trend of market demand and the characteristics of user groups, so as to better plan charging resources and service strategy;

[0053] Periodically receiving extreme weather warning information issued by the meteorological department enables the system to predict in advance the severe weather conditions that may affect the use of charging facilities and new energy vehicles, so as to take measures in advance;

[0054] Through real-time collection of charging pile operation data, faults and performance problems of charging piles can be found in time, the reliability and availability of equipment can be improved, and user waiting time and operation loss caused by equipment failure can be reduced;

[0055] Mastering new energy vehicle growth and battery capacity distribution data is helpful to optimize the layout and configuration of charging facilities, improve the utilization efficiency of charging resources, meet the growing market demand, and improve user experience and satisfaction;

[0056] Receiving extreme weather warning information in advance can help take timely protective measures, such as reinforcing charging facilities, adjusting charging strategies, etc., to reduce the damage risk of severe weather on charging facilities and new energy vehicles, and ensure the stable operation of the system and the safety of users;

[0057] It also includes a demand prediction module for receiving data from the perception module and obtaining demand prediction results after a prediction period based on the data from the perception module. The demand prediction results specifically include decomposing the prediction period into long-term, medium-term and short-term periods to obtain long-term demand prediction, medium-term demand prediction and short-term demand prediction. It should be noted that the time range of the three periods is set by real-time use, and the initial values are long-term (6-12 months), medium-term (1-6 months) and short-term (7-30 days);

[0058] It should be noted that the data received from the perception module includes charging pile operation data, new energy vehicle growth data, battery capacity distribution data and extreme weather warning information, etc. By comprehensively analyzing these data, the demand prediction results after the prediction period are obtained to provide decision basis for the dynamic optimization module;

[0059] Accurate demand prediction results help to reasonably plan the construction and upgrading of charging facilities, avoid over-investment or resource shortage, and improve investment efficiency and operational efficiency;

[0060] It can predict the trend of market demand in advance, so that the operating enterprise can adjust the operation strategy in advance, such as reasonably arranging the maintenance and repair plan of charging equipment, optimizing the charging price strategy, etc., to improve the competitiveness and profitability of the enterprise;

[0061] It provides accurate data support for the dynamic optimization module, so that it can better adjust and optimize in real time according to the changes in market demand, and improve the flexibility and adaptability of the entire system;

[0062] It also includes a dynamic optimization module for receiving data from the perception module and demand prediction results from the demand prediction module, and performing corresponding dynamic optimization based on the long-term demand prediction, medium-term demand prediction and short-term demand prediction in the demand prediction results.

[0063] It should be noted that the data from the perception module and the demand prediction results from the demand prediction module are combined with the actual inventory situation, and are comprehensively analyzed and evaluated to trigger corresponding inventory management operations, such as cross-regional allocation, emergency replenishment, etc., to ensure the normal operation of charging facilities and the satisfaction of market demand;

[0064] According to real-time data and prediction results, the inventory management strategy can be adjusted in time to ensure that the spare parts of the charging equipment are sufficient, reduce the downtime and maintenance delay of the equipment caused by insufficient inventory, and improve the reliability and availability of the equipment;

[0065] Through reasonable inventory management operations, inventory costs can be reduced, inventory accumulation and waste can be avoided, and the real-time adjustment and optimization functions of the dynamic optimization module can make the entire system better cope with market demand fluctuations and uncertainties, improve the stability and reliability of the system, and provide more stable and efficient charging services for users;

[0066] As an optional embodiment, the specific working steps of the perception module are as follows:

[0067] For real-time collection of charging pile operation data, the operation data of the charging pile are acquired in real time through sensors and data acquisition devices installed on the charging pile, including single-pile daily charging frequency Q and peak power duration T.

[0068] The specific acquisition steps include: first, the data acquisition device of the charging pile records the start and end time of each charging in real time. Whenever a vehicle starts charging, the system generates a charging record, including charging start time, end time, and other information. At the end of the statistical period, the system will count the number of charges Q of the charging pile in that day.

[0069] A power sensor is installed in the charging pile to monitor the power output in real time during the charging process. The control system of the charging pile records the power change data over time. When the power reaches the peak, the timing starts, and when the power drops below the peak, the timing stops, thereby obtaining the peak power duration T. It should be noted that by counting the charging frequency Q and the peak power duration T, the usage frequency and load of the charging pile can be understood. These data can provide a basis for the configuration and management of charging resources, such as reasonable arrangement of the number and layout of charging piles, optimization of charging price strategy, etc. According to the trend of charging frequency Q and peak power duration T, future charging demand can be predicted, and planning and construction of charging facilities can be done in advance, which helps to improve the utilization efficiency of charging resources, meet the growing demand for charging new energy vehicles, and promote the development of the new energy vehicle industry.

[0070] As an optional embodiment, the specific working steps of the perception module further include:

[0071] Through data sharing with the vehicle management system, the growth data of new energy vehicles within the range are dynamically acquired, and the vehicle battery capacity distribution data are acquired using the V2X communication protocol. It should be noted that the specific range within this range is set according to needs, and the initial value is 5KM.

[0072] The battery capacity distribution data specifically includes an average battery capacity V sales ; the average battery capacity V sales is obtained by summing up the battery capacities of all vehicles registered in the region and the total number of vehicles;

[0073] For periodic reception of extreme weather warning information issued by the meteorological department, specifically through the local meteorological department to obtain 72-hour extreme weather warning information; it should be noted that it specifically includes temperature, humidity, wind speed, and precipitation data.

[0074] As an optional embodiment, the specific working steps of the demand prediction module are:

[0075] For long-term period demand prediction, first, according to the obtained new energy vehicle sales data in the region in the past three long-term periods, normalize the new energy vehicle sales data in the three long-term periods to the interval [0, 1], select 64 LSTM units, add a Dropout layer to reduce overfitting, connect a Dense layer, and output the sales data of the next long-term period new energy vehicle, complete the construction of the model, input the normalized new energy vehicle sales data in the three long-term periods into the completed model, and output the new energy vehicle sales data P new in the next long-term period.

[0076] Then, according to the formula , the basic demand prediction Qh L of each long-term period is calculated; it should be noted that Qh L refers to the total demand for charging piles in the long term, which is predicted according to the sales growth of new energy vehicles, vehicle battery capacity, and power grid expansion, etc. The purpose is to predict future charging demand, evaluate market trends, and optimize resource allocation. Long-term demand Q L is an index that comprehensively considers the sales growth of new energy vehicles, vehicle battery capacity, and power grid expansion, etc., which is used to guide the long-term planning and construction of charging piles;

[0077] where Δ grid is the influence of power grid expansion; for ΔG grid , according to the power grid expansion planning data, determine the scale of power grid expansion in the future long term, including the capacity of expansion, the area of expansion, etc. According to the scale of power grid expansion and market conditions, calculate the influence of power grid expansion on charging demand ΔG grid , for example, if the power supply capacity of charging piles can be improved by 20% after power grid expansion, then ΔG grid = 0.2;

[0078] The power grid expansion can improve the power supply capacity of the charging pile, thereby increasing the charging demand. If the power grid is expanded, the charging power of the charging pile can be increased, and the charging time can be shortened, which will attract more new energy vehicle users to choose charging, thereby increasing the charging demand;

[0079] Then, each long-term period basic demand prediction Qh L is subtracted from the actual demand of each long-term period to obtain the historical residual error

[0080] The MLP neural network model is selected, all obtained historical residual errors

[0081] Then, the long-term period demand prediction Q is calculated according to the formula L It should be noted that by calculating the basic demand prediction, the demand of the future charging pile can be preliminarily estimated. Then, the historical residual error, that is, the difference between the actual charging demand and the basic prediction, is calculated. This step helps to identify the shortcomings of the basic model prediction. A residual prediction model is trained using the intermediate output of the basic model, external dynamic characteristics and time sequence derived characteristics. By inputting the related characteristics of the future time point into the trained residual model, the possible residual error in the future can be predicted, thereby dynamically correcting the basic demand prediction. Finally, the predicted residual error is added back to the basic demand prediction to obtain a more accurate and comprehensive total demand prediction of the charging pile. This series of steps not only improves the prediction accuracy, but also optimizes the resource allocation.

[0082] It should be noted that the proportion of new energy vehicles in the region has been changing in the long-term period. Along with the local policy or the price of new energy vehicles, the proportion of new energy vehicles in the region may change greatly in the long-term period. If the long-term demand cannot be predicted, the charging pile will be insufficient because the installation of the charging pile in the charging station requires time.

[0083] As an optional embodiment, the specific working steps of the demand prediction module further include:

[0084] For the medium-term period demand prediction, the medium-term demand Q

[0085]

[0086] is calculated according to the formula M , wherein p is the regional economic vitality factor, S seasonis a seasonal fluctuation coefficient, A is a charging station radiation capacity value, μ is a distance attenuation coefficient, and R is the number of charging piles within the charging station range; it should be noted that since the perception module is also used to dynamically obtain new energy vehicle growth data in the region, the above calculation is based on the initial charging pile where the perception module is located as the center, and the data within a 5km range around the center;

[0087] It should also be noted that through multi-dimensional data fusion and dynamic modeling, the prediction accuracy and practicality are significantly improved, providing strong support for the planning, operation and decision-making of the charging pile supply chain. By integrating multi-dimensional data such as economy, transportation and policy, breaking through the limitations of traditional single data sources, and through parameter dynamic adjustment and policy sensitivity modeling, the prediction model is self-optimized, the radiation effect and distance attenuation law of charging piles are quantified, and fine resource layout is supported.

[0088] where e -μd is a distance attenuation factor, used to describe the phenomenon that charging demand gradually decreases with increasing distance, d is the distance attenuation coefficient of the region where the charging station is located, and the value is calculated by comparing the total charging amount of the last quarter of each region in the region with the total charging amount of the region where the charging station is located. The ratio represents how many times the charging demand of the region where the charging station is located is compared to other regions. The ratio is taken as the natural logarithm, which reflects the exponential decay relationship of charging demand with distance. The natural logarithm result is divided by the edge distance, μ represents the change rate of charging demand per unit distance, i.e. for every additional kilometer, the charging demand will decrease by a certain percentage. It should be noted that the greater the speed μ at which charging demand decreases with increasing distance, the faster the charging demand decreases with distance; otherwise, it decreases slowly.

[0089] It should be noted that for large areas, the demand of each region in the region is different, although charging stations will be installed in each region, but the demand of some regions, such as parking lots, is greater, and more charging piles are needed. If fixed charging piles are installed in each region, the cost will increase, so a small number of mobile energy storage charging piles need to be configured, which can be moved to the appropriate area to meet the use demand in special circumstances, and the demand prediction of the medium-term period in the region can be obtained. It can better serve as a reference for the subsequent allocation of mobile energy storage charging piles.

[0090] As an optional embodiment, the specific working steps of the demand prediction module further include:

[0091] For the regional economic vitality factor p, is obtained, wherein 1 is the annual GDP of the city where the region is located, 2 is the annual total GDP of the country, N1 is the number of newly registered cars in the region during the period, and N2 is the total number of cars in the region; it should be noted that the economic vitality and the popularization degree of new energy vehicles in the region are reflected, so as to predict the potential of the charging market;

[0092] For the seasonal fluctuation coefficient S season , based on the fitting of the charging amount data in the same period in the past three years, specifically, is obtained, wherein t represents the month; it should be noted that the seasonal change of the charging demand is captured, and the accuracy of the prediction is improved;

[0093] For the charging station radiation capacity value A, according to the formula is obtained, wherein P t is the daily average passenger flow in the charging station range, D p is the charging pile density, and N b is the number of public transportation lines near the region. It should be noted that P t is the daily average passenger flow in the charging station range, which is within 5km around the charging pile, and N is the number of public transportation lines near the region, which is specifically the number of public transportation lines within 1km near the region.

[0094] As an optional embodiment, the specific working steps of the demand prediction module further include:

[0095] For the demand prediction of the short-term period, the specific method is:

[0096]

[0097] , wherein ΔT is the difference between the current temperature and the standard temperature, R r is the current rainfall intensity, W S is the current wind speed, ΔC is the battery capacity attenuation rate, which is obtained by subtracting the current battery capacity from the initial battery capacity and then dividing by the initial battery capacity, X is the traffic congestion value, which is obtained by , wherein TL is the traffic index, is the benchmark value of the charging demand, that is, the basic level of the charging demand without other influencing factors; it should be noted that the prediction accuracy is improved:

[0098] By taking into account real-world factors such as weather, battery degradation, and traffic conditions, the formula can more accurately reflect changes in charging demand. Each factor is adjusted according to current conditions, enabling the forecast to respond to environmental changes in real time. More accurate demand forecasting helps optimize the use of charging infrastructure, ensuring sufficient charging resources in high-demand times and locations. By avoiding overcharging at night during low-demand periods and ensuring sufficient capacity during high-demand periods, the return on investment can be improved, and sufficient charging capacity can be ensured in the event of demand surges such as severe weather or traffic congestion.

[0099] It should be noted that traffic indexes can be obtained through map software.

[0100] As an optional embodiment, the specific working steps of the dynamic optimization module include the following:

[0101] At each interval, the forecast error E of the short-term demand forecast is calculated, specifically according to the formula: The prediction error E is calculated, where Q1 is the actual charging demand value. If E is greater than 15%, then an adjustment is made. The weighting coefficients in the model are 0.3, 0.3, and 0.4. When E is greater than 0, these coefficients are increased by 2%; if E is less than 0, they are decreased by 2%. It should be noted that through an error feedback mechanism, the model can monitor the prediction error in real time and dynamically adjust the weighting coefficients based on the error. When the prediction error continues to exceed a certain threshold, the model will trigger an automatic update to ensure the accuracy and reliability of the prediction. This method helps improve the accuracy of charging demand forecasting and optimize the operation and management of charging stations.

[0102] As an optional embodiment, the specific working steps of the dynamic optimization module further include the following:

[0103] For long-term periods, according to the formula The number of newly added charging piles N is calculated, and based on the number of newly added charging piles N, the number of charging piles with a total quantity of N to be installed in this area during the next long-term period is calculated, where P x T1 represents the power of a single charging station, and T1 represents the average working time of the charging station. It should be noted that the average working time of the charging station is generally the working time of the area, which is 8 hours.

[0104] During the medium-term period, the real-time energy reserve data E and safe energy threshold E1 of each charging pile in the region are obtained from the sensing module, and the demand forecast Q for each charging pile in the short-term, medium-term, and long-term periods are obtained from the demand forecast module. L Q M and Q S ;

[0105] The inventory monitoring degree index H of each area is calculated, and when the inventory monitoring degree index H is greater than a preset threshold, the movable charging pile is moved to the area;

[0106] In a short-term period, the dynamic time-sharing pricing N of the charging pile in the short-term period is calculated according to the formula N j = P L *(1+0.1*Q L ),wherein P is the basic electricity price. j L

[0107] It should be noted that through the optimization measures of long-term, medium-term and short-term, the changes in the number of new energy vehicles in the area, the demand differences of different areas and the actual influencing factors can be comprehensively responded to, and the reasonable allocation and efficient use of the charging pile are ensured;

[0108] According to the long-term demand prediction, the number of new charging piles is determined, and the corresponding number of charging piles is installed in the next long-term period; according to the medium-term demand prediction, the power reserves and demand of the charging piles in each area in the area are monitored in real time, and the distribution of the charging piles is dynamically adjusted;

[0109] According to the short-term demand prediction, the dynamic time-sharing pricing of the charging pile is calculated, and the user is guided to reasonably use the charging pile in different time periods;

[0110] The utilization rate of the charging pile can be improved, the investment cost can be reduced, and sufficient charging resources can be ensured at the time and place of high demand, and the user experience is improved.

[0111] As an optional embodiment, the specific steps of calculating the inventory monitoring degree index H of the charging pile are:

[0112] The inventory monitoring degree index H of each charging pile is calculated by H=(E-E1)+(Q L *0.2)+(Q M *0.1)+(Q S *0.02). It should be noted that by considering the predicted charging demand in short-term, medium-term and long-term, the power resources can be reasonably allocated. In the demand peak period, the power reserve can be increased in advance to avoid charging interruption due to insufficient power, and in the demand trough period, the power reserve can be reduced to reduce the operating cost. By optimizing the power reserve, the resource utilization efficiency of the charging pile can be improved, the power waste can be reduced, the charging service continuity of the charging pile can be ensured by timely supplementing the power reserve, the user satisfaction can be improved, the waiting time of the user due to insufficient power of the charging pile can be reduced, and the user experience can be improved.

[0113] Working principle ​

[0114] The perception module is used to collect operation data of the charging pile in real time, and specifically includes single-pile daily average charging times and peak power duration;

[0115] The perception module is also used to dynamically obtain new energy vehicle growth data in the region and obtain vehicle battery capacity distribution data;

[0116] It is also used to periodically receive extreme weather warning information issued by the meteorological department, and the perception module transmits all data;

[0117] It also includes a demand prediction module, which is used to receive data of the perception module and obtain demand prediction results after a prediction period according to the data of the perception module. The demand prediction results specifically include decomposing the prediction period into three time periods of long-term, medium-term and short-term, obtaining long-term demand prediction, medium-term demand prediction and short-term demand prediction;

[0118] It also includes a dynamic optimization module, which is used to receive data of the perception module and demand prediction results of the demand prediction module, and perform corresponding dynamic optimization according to long-term demand prediction, medium-term demand prediction and short-term demand prediction in the above demand prediction results.

[0119] The above is only the preferred embodiment of the present application, and the protection scope of the present application is not limited to the above-mentioned embodiments only. Any technical solution falling within the scope of the present application should be considered as falling within the protection scope of the present application. It should be noted that for ordinary technical personnel in the technical field, some improvements and refinements without departing from the principles of the present application should also be considered as falling within the protection scope of the present application.

Claims

1. An artificial intelligence based supply chain demand forecasting and dynamic optimization system, characterized in that, The perception module is used to collect operation data of the charging pile in real time, and specifically includes single-pile daily charging frequency and peak power duration; The perception module is also used to dynamically obtain new energy vehicle growth data in the region and obtain vehicle battery capacity distribution data; It is also used to periodically receive extreme weather warning information issued by the meteorological department, and the perception module transmits all data; It also includes a demand prediction module, which is used to receive data from the perception module and obtain demand prediction results after a prediction period based on the data from the perception module. The demand prediction results specifically include dividing the prediction period into long-term, medium-term and short-term periods to obtain long-term demand prediction, medium-term demand prediction and short-term demand prediction. It also includes a dynamic optimization module, which is used to receive data from the perception module and demand prediction results from the demand prediction module, and perform corresponding dynamic optimization based on the long-term demand prediction, medium-term demand prediction and short-term demand prediction in the demand prediction results. The specific working steps of the demand prediction module are: For the demand prediction of long-term period, firstly, according to the obtained new energy vehicle sales data in the region in the past three long-term periods, the new energy vehicle sales data in the three long-term periods are normalized to the interval [0, 1], 64 LSTM units are selected, a Dropout layer is added to reduce overfitting, a Dense layer is connected, and the output of the next long-term period new energy vehicle sales data is output, the model is completed, and the normalized new energy vehicle sales data in the three long-term periods is input into the completed model, and the new energy vehicle sales data P of the next long-term period is output new ; The base demand forecast Qh for each long-term period is then calculated according to the formula L where ΔG grid is the grid expansion impact;​ Then each long-term period base demand forecast Qh L is subtracted from the actual demand for each long-term period to obtain the historical residual ∈1; Select the MLP neural network model, divide all obtained historical residual data into training set, validation set and test set, divide according to the proportion of 70%, 15% and 15%, train to obtain the residual model, and input the historical residual into the residual model to obtain the residual of the next long-term period; The long-term period demand forecast Q is then calculated according to the formula Q = Q + Q L ; The specific working steps of the demand prediction module also include: For medium-term demand prediction, according to the formula The medium-term demand Q is calculated M wherein ρ is the regional economic vitality factor, S season is the seasonal fluctuation coefficient, A is the charging station radiation capacity value, μ is the distance attenuation coefficient, and R is the number of charging piles within the range of the charging station. where e -μd is a distance decay factor used to describe the phenomenon that the charging demand gradually weakens with the increase of distance, d is a distance decay coefficient of the area where the charging station is located; The specific working steps of the demand prediction module also include: For short-term demand prediction, the specific method is; where ΔT is the difference between the current temperature and the standard temperature, R r is the current rainfall intensity, W S is the current wind speed, ΔC is the battery capacity decay rate, obtained by subtracting the current battery capacity from the initial battery capacity and dividing by the initial battery capacity, X is the traffic congestion value, obtained by where TL is the traffic index, is the baseline value of the charging demand, i.e. the basic level of the charging demand in the absence of other influencing factors.

2. The system, as claimed in claim 1, wherein, The specific working steps of the perception module are as follows: For real-time collection of operation data of the charging pile, the operation data of the charging pile is obtained in real time through sensors and data collection equipment installed on the charging pile, including single-pile daily charging frequency Q and peak power duration T; The specific acquisition steps include: first, the data collection equipment of the charging pile records the start and end time of each charging in real time. Whenever a vehicle starts charging, the system will generate a charging record, including charging start time and end time information. At the end of the statistical period, the system will count the number of charges Q of the charging pile in that day. Install a power sensor in the charging pile to monitor the power output in real time during the charging process. The control system of the charging pile will record the power change data over time. When the power reaches the peak, start timing, and stop timing when the power drops below the peak, so as to obtain the duration T of the peak power.

3. The system, as claimed in claim 1, wherein, The specific working steps of the perception module also include: Through data sharing with the vehicle management system, the growth data of new energy vehicles in the range is dynamically obtained, and at the same time, the vehicle battery capacity distribution data is obtained by using the V2X communication protocol. The battery capacity distribution data specifically includes an average battery capacity V sales ; the average battery capacity V sales is obtained by summing the battery capacities of all vehicles registered in the area / the total number of vehicles For periodically receiving extreme weather warning information issued by the meteorological department, 72-hour extreme weather warning information is obtained through the local meteorological department.

4. The system, as claimed in claim 1, wherein, The specific working steps of the demand prediction module also include: For regional economic vitality factor p, is obtained, wherein G1 is the annual GDP of the city where the region is located, G2 is the annual total GDP of the country, N1 is the number of newly registered cars in the city where the region is located within a period of time, and N2 is the total number of cars in the city where the region is located. For the seasonal fluctuation coefficient S season , obtained based on the fitting of the charging amount data of the same period in the past three years, specifically, is obtained, where t represents the month. For the charging station radiation capacity value A, according to the formula is obtained, wherein P t is the daily average passenger flow in the charging station range, D p is the charging pile density, N b is the number of bus lines near the region.

5. The artificial intelligence based supply chain demand forecasting and dynamic optimization system, as claimed in claim 1 wherein, The specific working steps of the dynamic optimization module include the following: Every other cycle, the prediction error E of the short-term period demand prediction is calculated, specifically according to the formula: The prediction error E is calculated and obtained, wherein Q1 is the actual charging demand value, if E is greater than 15%, then adjust The weight coefficients 0.3, 0.3 and 0.4 in the formula are adjusted to increase 2% when E is greater than 0, and to decrease 2% when E is less than 0.

6. The artificial intelligence based supply chain demand forecasting and dynamic optimization system as claimed in claim 5 wherein, The specific working steps of the dynamic optimization module further include the following: For a long period, the newly added charging pile number N is calculated according to the formula According to the newly added charging pile number N, the number of charging piles to be installed in the next long period in the region is N, wherein P x is the power of a single charging pile, and T1 is the average working time of the charging pile. In the medium term, the real-time electric energy reserve data E and the safe electric energy threshold E1 of each charging pile in the region are obtained from the perception module, and the short-term, medium-term and long-term demand prediction Q of each charging pile is obtained from the demand prediction module L , Q M , and Q S ; The inventory monitoring degree index H of each region is calculated, and when the inventory monitoring degree index H is greater than a preset threshold, the movable charging pile is moved to the region. In the short term, according to formula N j =P L *(1+0.1*Q L The dynamic time-of-use pricing N for charging piles in the short term is calculated. j , where P L Based on the basic electricity price.

7. The artificial intelligence based supply chain demand forecasting and dynamic optimization system as claimed in claim 6 wherein, The specific steps of calculating the inventory monitoring degree index H of the charging pile are as follows: The inventory monitoring index H of each charging pile is calculated by H = (E - E1) + (Q L *0.2) + (Q M *0.1) + (Q S *0.02).

Images