Charging station operation intelligent pricing method, operation server and storage medium

Abstract

The invention relates to the technical field of charging stations, and discloses a charging station operation intelligent pricing method, an operation server and a storage medium. The method comprises the following steps: acquiring user historical charging order original data of a charging station; performing deep analysis on the original data of the historical charging order of the user by using a data mining algorithm, extracting key factors influencing pricing, and determining the influence degree of the key factors on intelligent pricing; based on the influence degree of the key factors on intelligent pricing and original data of historical charging orders of the user, a data fusion model is used to generate multiple intelligent pricing reasoning fusion data with reasoning trajectories; and generating a charging station operation intelligent pricing strategy based on the pricing fusion model and the plurality of intelligent pricing reasoning fusion data with the reasoning tracks. Therefore, a charging station operation intelligent pricing strategy can be formulated according to the original data of the historical charging orders of the user, the charging price of each time period is automatically adjusted, and greater economic value is created for the charging station.

Description

Technical Field

[0001] The present invention relates to the technical field of charging stations, and particularly to a method for intelligent pricing of charging station operations, an operation server, and a storage medium.

Background Art

[0002] With the rapid expansion of the electric vehicle market, charging stations, as key infrastructure, are facing increasingly complex operation challenges. In a charging station, multiple charging piles are generally deployed, and the charging piles include fast charging piles and / or slow charging piles. The charging piles can charge electric vehicles. In large-scale charging stations, the charging price pricing strategy is generally determined according to the traditional time-of-use electricity price. The traditional division of peak hours, flat peak hours, and off-peak hours is to divide each day into three periods, with each period being preferably 8 hours. The peak hours of each day are 7:00 - 11:00 and 19:00 - 23:00; the normal hours are 11:00 - 19:00; the off-peak hours are 23:00 - 7:00 the next day. For different charging periods, different charging prices are implemented, and the same charging price is implemented during the same period. This charging price pricing strategy does not fully consider the key factors affecting charging pricing in the charging station, nor does it consider the grid load pressure of the charging station and the overall economic benefits of the charging station, resulting in an unreasonable charging price structure and charging price pricing strategy that are likely to cause concentrated charging behaviors of users and unreasonable distribution of charging demands, which not only increases the grid load pressure but also reduces the overall economic benefits of the charging station.

Summary of the Invention

[0003] Embodiments of the present invention aim to provide a method for intelligent pricing of charging station operations, an operation server, and a storage medium, aiming to solve the problem of unreasonable charging price pricing in current charging stations, which causes concentrated charging behaviors of users and unreasonable distribution of charging demands.

[0004] To solve the above technical problems, a first aspect embodiment of the present invention provides a method for intelligent pricing of charging station operations, which is applied to an operation server. The method for intelligent pricing of charging station operations includes:

[0005] Obtain the original data of the user's historical charging orders in the charging station;

[0006] Use a data mining algorithm to process the original data of the user's historical charging orders, extract the key factors affecting pricing, and determine the influence degree of the key factors on intelligent pricing;

[0007] Based on the influence degree of the key factors on intelligent pricing and the original data of the user's historical charging orders, use a data fusion model to generate multiple intelligent pricing inference fusion data with inference trajectories;

[0008] Generate an intelligent pricing strategy for charging station operation based on a pricing fusion model and multiple intelligent pricing inference fusion data with inference trajectories.

[0009] Optionally, the use of data mining algorithms to process the original user historical charging order data, extract key factors affecting pricing, and determine the degree of influence of key factors on intelligent pricing includes:

[0010] Cluster the original user historical charging order data according to charging time using the K-means clustering algorithm, and divide different charging periods of a day into peak periods, flat peak periods, and off-peak periods;

[0011] Use the Pearson correlation coefficient to analyze the correlation between charging income and key factors including the number of orders, charging duration, and charging power, and determine the degree of influence of each key factor on income;

[0012] Use the XGBoost algorithm to quantify the degree of influence of each key factor on intelligent pricing.

[0013] Optionally, the clustering of the original user historical charging order data according to charging time using the K-means clustering algorithm, and dividing different charging periods of a day into peak periods, flat peak periods, and off-peak periods includes:

[0014] Extract each charging period of charging time in hours from the original user historical charging order data, and initially divide different charging periods into peak periods, flat peak periods, and off-peak periods;

[0015] Extract features related to the charging period from the original user historical charging order data: charging start time, charging end time, charging power, charging duration;

[0016] Use the K-means clustering algorithm to cluster the extracted features related to the charging period, and finally divide different charging periods of a day into peak periods, flat peak periods, and off-peak periods.

[0017] Optionally, based on the degree of influence of the key factors on intelligent pricing and the original user historical charging order data, the use of a data fusion model to generate multiple intelligent pricing inference fusion data with inference trajectories includes:

[0018] In the data generation stage, optimize the thought chain prompt words, construct multiple questions, and use a large language model to generate multiple answers for each question;

[0019] In the data filtering stage, use a generative artificial intelligence model to automatically screen the multiple answers generated by the large language model for each question, and retain several intelligent pricing inference fusion data with inference trajectories after screening.

[0020] Optionally, the optimized chain-of-thought prompt words include:

[0021] Embedding explicit generation answer format instructions in the chain-of-thought prompt words;

[0022] Fusing the influence degree of the key factors on intelligent pricing with the original data of the user's historical charging orders, and injecting the fused data into the chain-of-thought prompt words as prior knowledge.

[0023] Optionally, in the data filtering stage, using a generative artificial intelligence model to automatically screen multiple answers generated by a large language model for each question, and retaining several intelligent pricing inference fusion data with inference tracks after screening, including:

[0024] Automatically screening multiple answers generated by the large language model for each question according to preset rules and preset output format specifications, and screening out answers that do not conform to the preset rules and preset output format specifications;

[0025] Using a generative artificial intelligence model to re-evaluate the answers that do not conform to the preset rules and preset output format specifications, accurately identifying the answers that are misjudged as wrong but are actually correct, and finally retaining several intelligent pricing inference fusion data with inference tracks after screening.

[0026] Optionally, the preset rules include that the adjustment range of the energy fee is within the first range, the adjustment ranges of the start-up fee, the detention fee, and the duration fee are within the second range, and the monthly charging income prompt is controlled within the third range; the preset output format specifications include <think>Inference process< / think> The output format of the inference result.

[0027] Optionally, generating an intelligent pricing strategy for charging station operation based on the pricing fusion model and the multiple intelligent pricing inference fusion data with inference tracks includes:

[0028] Using a sequence prediction model to obtain key knowledge in the field of intelligent pricing;

[0029] Selecting DeepSeek R1 as the base model, and using the large model fine-tuning technology to fine-tune the combination of the key knowledge in the field of intelligent pricing output by the sequence prediction model and the several intelligent pricing inference fusion data with inference tracks retained after screening to generate a high-quality intelligent pricing strategy for charging station operation.

[0030] Correspondingly, the second aspect embodiment of the present invention provides an operation server, including: a memory, a processor, and a computer program stored on the memory and running on the processor, where the computer program, when executed by the processor, implements the method for intelligent pricing of charging station operation in the first aspect embodiment of the present invention.

[0031] Correspondingly, an embodiment of the third aspect of the present invention provides a storage medium, on which a program of a method for intelligent pricing of charging station operation is stored. When the program of the method for intelligent pricing of charging station operation is executed by a processor, it implements the method for intelligent pricing of charging station operation described in the embodiment of the first aspect of the present invention.

[0032] Compared with the prior art, a method for intelligent pricing of charging station operation, an operation server and a storage medium provided by the embodiments of the present invention. The method for intelligent pricing of charging station operation is applied to the operation server, and includes: obtaining the original data of the user's historical charging orders of the charging station; using a data mining algorithm to deeply analyze the original data of the user's historical charging orders, extracting the key factors affecting pricing, and determining the influence degree of the key factors on intelligent pricing; based on the influence degree of the key factors output by the data mining algorithm on intelligent pricing and the original data of the user's historical charging orders, using a data fusion model to generate multiple intelligent pricing inference fusion data with inference tracks; based on the pricing fusion model and multiple intelligent pricing inference fusion data with inference tracks, generating an intelligent pricing strategy for charging station operation. Thereby, an intelligent pricing strategy for charging station operation can be formulated according to the original data of the user's historical charging orders, and the charging price of each time period can be automatically adjusted according to the intelligent pricing strategy for charging station operation, so as to guide the reasonable distribution of the user's charging demand, improve the utilization efficiency of the charging equipment of the charging station, and ultimately create greater economic value for the charging station operator, achieving the balance between the maximization of the charging station's revenue and the reasonable allocation of resources. Thus, the problem that the charging price of the current charging station is unreasonably priced, resulting in the concentration of user charging behavior and the unreasonable distribution of charging demand, can be solved.

BRIEF DESCRIPTION OF THE DRAWINGS

[0033] One or more embodiments are exemplarily illustrated by the pictures in the corresponding drawings. These exemplary illustrations do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings are represented as similar elements, unless otherwise stated, and the drawings in the drawings do not constitute a proportional limitation.

[0034] Figure 1 is a schematic diagram of the composition of a charging station provided by the present invention;

[0035] Figure 2 is a schematic flow diagram of a method for intelligent pricing of charging station operation provided by the present invention;

[0036] Figure 3 is a schematic structural diagram of an operation server provided by the present invention.

DETAILED DESCRIPTION

[0037] To facilitate the understanding of the present invention, the present invention will be described in more detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that when an element is described as "fixed to" another element, it can be directly on the other element, or there can be one or more intermediate elements therebetween. When an element is described as "electrically connected to" another element, it can be directly connected to the other element, or there can be one or more intermediate elements therebetween. The terms "upper", "lower", "inner", "outer", "bottom", etc. used in this specification indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be construed as a limitation to the present invention. In addition, the terms "first", "second", "third", etc. are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0038] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not used to limit the present invention. The term "and / or" used in this specification includes any and all combinations of one or more of the related listed items.

[0039] In addition, the technical features involved in different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0040] In one embodiment, as Figure 1 shown, the charging station 100 includes: an operation server 10, a plurality of charging piles 20, a power distribution system 30, a photovoltaic energy storage system 40, a plurality of charging spaces 50, and auxiliary facilities 60; where:

[0041] The plurality of charging piles 20 include a number of fast-charging piles and / or a number of slow-charging piles. The fast-charging piles are DC charging piles with a power of 60 - 350 kW, which are used for the rapid charging of electric vehicles and can charge to 80% in 30 minutes; the slow-charging piles are AC charging piles with a power of 7 - 22 kW, which are used for charging electric vehicles parked for a long time.

[0042] The power distribution system 30 includes a transformer and a power distribution cabinet; the transformer is used to convert the high-voltage mains into a voltage available for the charging piles, for example, converting the 10 kV high-voltage mains into the 380 V available voltage for the charging piles; the power distribution cabinet is used for power distribution and protection.

[0043] The photovoltaic energy storage system 40 includes solar panels and energy storage batteries, which are used to achieve peak shaving and valley filling and reduce the electricity cost.

[0044] The charging parking space 50 is used for parking electric vehicles to be replenished with energy.

[0045] The operation server 10 is communicatively connected to each charging pile 20, and is used to control the charging status of each charging pile 20 and perform load balancing on these charging piles 20, configure intelligent pricing for the operation of the charging station, and communicate with the cloud service platform through a wired network (such as a fiber optic network) and / or a wireless network (such as a 4G / 5G wireless network), and upload charging data to the cloud service platform in real time.

[0046] The auxiliary facilities 60 include an intelligent gate with license plate recognition, monitoring equipment, etc.

[0047] With the rapid expansion of the electric vehicle market, charging stations, as key infrastructure, are facing increasingly complex operation challenges. In large-scale charging stations, the charging price pricing strategy is generally determined based on the traditional time-of-use electricity price. The traditional peak period, flat peak period, and off-peak period are divided by dividing each day into three periods, with each period being preferably 8 hours. The peak period of each day is 7:00 - 11:00 and 19:00 - 23:00; the normal period is 11:00 - 19:00; the off-peak period is 23:00 - 7:00 the next day. For different charging periods, different charging prices are implemented, and the same charging price is implemented during the same period. This charging price pricing strategy does not fully consider the key factors affecting charging pricing in the charging station, nor does it consider the grid load pressure of the charging station and the overall economic benefits of the charging station, resulting in an unreasonable charging price structure. The charging price pricing strategy is likely to cause concentrated user charging behavior and unreasonable charging demand distribution, which not only increases the grid load pressure but also reduces the overall economic benefits of the charging station.

[0048] In view of this, as Figure 2 shown, the present invention provides a method for realizing intelligent pricing for the operation of a charging station based on data mining and integrating the large model fine-tuning algorithm, which is applied to the operation server. The method includes:

[0049] S1. Obtain the original data of the user's historical charging orders in the charging station;

[0050] S2. Use the data mining algorithm to process the original data of the user's historical charging orders, extract the key factors affecting pricing, and determine the influence degree of the key factors on intelligent pricing;

[0051] S3. Based on the influence degree of the key factors output by the data mining algorithm on intelligent pricing and the original data of the user's historical charging orders, use the data fusion model to generate multiple intelligent pricing inference fusion data with inference trajectories;

[0052] S4. Generate an intelligent pricing strategy for the operation of the charging station based on the pricing fusion model and multiple intelligent pricing inference fusion data with inference trajectories.

[0053] In this embodiment, a method for realizing intelligent pricing of charging station operation by providing a fine-tuning algorithm for data mining fusion large model is applied to the operation server, including: obtaining the original data of the user's historical charging orders of the charging station; using the data mining algorithm to deeply analyze the original data of the user's historical charging orders, extracting the key factors affecting pricing, and determining the influence degree of the key factors on intelligent pricing; based on the influence degree of the key factors output by the data mining algorithm on intelligent pricing and the original data of the user's historical charging orders, using the data fusion model to generate multiple intelligent pricing inference fusion data with inference trajectories; generating an intelligent pricing strategy for the operation of the charging station based on the pricing fusion model and multiple intelligent pricing inference fusion data with inference trajectories. Thus, an intelligent pricing strategy for the operation of the charging station can be formulated according to the original data of the user's historical charging orders, and the charging price of each period can be automatically adjusted according to the intelligent pricing strategy for the operation of the charging station, so as to guide the reasonable distribution of the user's charging demand, improve the utilization efficiency of the charging equipment of the charging station, and ultimately create greater economic value for the charging station operator, achieving the balance between the maximum revenue of the charging station and the reasonable allocation of resources. Thus, the problem of unreasonable charging price pricing of the current charging station causing concentrated user charging behavior and unreasonable charging demand distribution can be solved.

[0054] In one embodiment, in step S1, obtain the original data of the user's historical charging orders of the charging station.

[0055] The original data of the user's historical charging orders includes the user's charging detail data, specifically including the start-up fee, charging start time, charging end time, duration fee, charging power, charging duration, energy fee, and detention fee.

[0056] The start-up fee refers to a fixed fee that will be charged in advance when the electric vehicle starts charging, regardless of the final charging amount. The purpose of setting the start-up fee is to cover the basic operating costs of the charging pile (such as equipment startup, network communication, etc.), and to avoid losses to the charging station operator caused by users' short-term charging (such as only charging for 5 minutes).

[0057] The duration fee, also known as the charging duration fee, refers to the fee charged by the charging operator of the charging station according to the total time that the user occupies the charging pile (including the charging time and the staying time after charging is completed), rather than simply charging according to the actual charging amount (kWh). This charging method is common in scenarios with high requirements for the utilization rate of charging pile resources.

[0058] The charging power refers to the charging degree of the electric vehicle.

[0059] The charging duration refers to the charging time of an electric vehicle from the start of charging to the end of charging.

[0060] The energy cost refers to the electricity cost when an electric vehicle is charging, which is billed according to the actual charging amount (kilowatt-hours, kWh).

[0061] The detention fee, also known as the occupancy fee or overtime occupancy fee, refers to the additional fee generated when an electric vehicle fails to leave the charging space in time within the specified free grace period after charging is completed, resulting in the charging pile being occupied for a long time. The core purpose of setting the detention fee is to improve the turnover rate of the charging pile, avoid vehicles that are fully charged from occupying resources for a long time, and affect the charging of other users.

[0062] In this embodiment, through the original data of the user's historical charging orders in the charging station, it is possible to provide the original data of the user's historical charging orders for in-depth analysis using data mining algorithms, so as to extract the key factors affecting pricing and formulate an intelligent pricing strategy for the operation of the charging station.

[0063] In one embodiment, in step S2, a data mining algorithm is used to process the original data of the user's historical charging orders, extract the key factors affecting pricing, and determine the degree of influence of the key factors on intelligent pricing. Specifically, it includes:

[0064] S21. Cluster analysis: The original data of the user's historical charging orders is clustered according to the charging time using the K-means clustering algorithm, and different charging periods of a day are divided into peak periods, flat peak periods, and off-peak periods.

[0065] K-means is an unsupervised learning algorithm used to divide data into K non-overlapping clusters, so that the data points within the same cluster are as similar as possible, while the data points between different clusters are as different as possible. Its core idea is to find the center (centroid) of the cluster through iterative optimization and assign the data points to the cluster to which the nearest centroid belongs.

[0066] The original data of the user's historical charging orders is clustered according to the charging time using the K-means clustering algorithm, and different charging periods of a day are divided into peak periods, flat peak periods, and off-peak periods. Specifically, it includes:

[0067] S211. Initially divide the peak period, flat peak period, and off-peak period: Extract each charging period of the charging time from the original data of the user's historical charging orders in hours; initially divide the different charging periods into: peak period, the charging time period is from 7:00 to 11:00 and from 19:00 to 23:00; flat peak period, the charging time period is from 11:00 to 19:00; off-peak period, the charging time period is from 23:00 to 7:00.

[0068] S212. Clustering Feature Selection: Extract features related to the charging period from the original data of the user's historical charging orders: charging start time, charging end time, charging power, and charging duration.

[0069] S213. Clustering Implementation: Use the K-means clustering algorithm to cluster the features related to the charging period, and finally divide the different charging periods of a day into peak periods, flat peak periods, and off-peak periods.

[0070] For example, after using the K-means clustering algorithm to cluster the features related to the time period, it is obtained that:

[0071] In the time period from 0:00 to 7:00, the clustering center hour is 3.4, so the time period from 0:00 to 7:00 is divided into the off-peak period;

[0072] In the time period from 7:00 to 17:00, the clustering center hour is 15.5, so the time period from 7:00 to 17:00 is divided into the flat peak period;

[0073] In the time period from 17:00 to 23:00, the clustering center hour is 19.8, so the time period from 17:00 to 23:00 is divided into the peak period.

[0074] S22. Correlation Analysis: Use the Pearson Correlation Coefficient to analyze the correlation between charging income and key factors such as the number of orders, charging duration, and charging power, and determine the influence degree of each key factor on income, so as to help the operator identify the factors that have the most significant impact on income and optimize the pricing strategy and resource allocation.

[0075] The Pearson correlation coefficient represents a measure of the linear correlation degree between two continuous variables, and its value range is [-1, 1], where: 1 represents a perfect positive correlation, -1 represents a perfect negative correlation, and 0 represents no linear correlation. The Pearson correlation coefficient formula is as follows:

[0076]

[0077] Among them, r is the Pearson correlation coefficient, X and Y are variables, is the mean value.

[0078] After obtaining the original data of the user's historical charging orders at a charging station for a period of time, extract the data of key factors such as charging income, number of orders, charging duration, and charging power from the original data of the user's historical charging orders, and clean the data of the extracted key factors. For example, eliminate abnormal data with a charging duration of 0 or a negative income.

[0079] Perform data standardization processing on the data of the key factors after cleaning.

[0080] Calculate the correlation coefficients between each key factor and the charging income using the Pearson correlation coefficient formula, obtain the correlation coefficients between each key factor and the charging income, and determine the influence degree of each key factor on the charging income.

[0081] For example, the correlation coefficients between each key factor and the charging income calculated using the Pearson correlation coefficient formula are as follows:

[0082] The correlation coefficient r between the charging power and the charging income is 0.9, then it is determined that there is a strong positive correlation between the charging power and the charging income. At this time, increasing the single - charge power can directly increase the charging income.

[0083] The correlation coefficient r between the order quantity and the charging income is 0.6, then it is determined that there is a medium positive correlation between the order quantity and the charging income. At this time, increasing the number of users or the repurchase rate is the key to increasing the charging income.

[0084] The correlation coefficient r between the charging duration and the charging income is 0.2, then it is determined that there is a weak or no correlation between the charging duration and the charging income. At this time, it shows that the charging duration has no direct impact on the charging income.

[0085] S23. Key factor analysis: Use the XGBoost algorithm to quantify the influence degree of each key factor on intelligent pricing, and provide data support for intelligent pricing.

[0086] The XGBoost algorithm can automatically handle non - linear relationships, sort feature importance, and output feature importance, which can quantify the influence degree of each key factor on the charging income.

[0087] Using the XGBoost algorithm to quantify the influence degree of key factors on intelligent pricing specifically includes:

[0088] Use the XGBoost algorithm to process the correlation coefficients between each key factor and the charging income calculated by the Pearson correlation coefficient formula, output the feature importance of each key factor, and quantify the influence degree of each key factor on intelligent pricing.

[0089] For example: After using the XGBoost algorithm to process the correlation coefficients of the order quantity, charging duration, and charging power calculated by the Pearson correlation coefficient formula, output the feature importance of each key factor of the order quantity, charging duration, and charging power, and quantify the influence degree of each key factor on intelligent pricing, as follows:

[0090] Charging power (kWh), the feature importance is 0.35, which indicates that the charging power directly affects the income, and increasing the single - charge amount can significantly increase the charging income.

[0091] For the order quantity, with a feature importance of 0.25, it indicates that increasing the number of users or the repurchase rate is the key to increasing charging revenue;

[0092] For the charging duration, with a feature importance of 0.10, it indicates that the charging duration has no direct impact on charging revenue.

[0093] In the prior art, the traditional division basis for peak hours, off-peak hours, and valley hours is to divide each day into three time periods, with each period being preferably 8 hours. The peak hours of each day are from 7:00 to 11:00 and from 19:00 to 23:00; the normal hours are from 11:00 to 19:00; and the valley hours are from 23:00 to 7:00 of the next day. This traditional way of dividing charging time periods does not fully consider the key factors affecting charging pricing in the charging station, nor does it consider the grid load pressure of the charging station and the overall economic benefits of the charging station.

[0094] In this embodiment, by clustering the original data of users' historical charging orders according to the charging time using the K-means clustering algorithm, different charging time periods of a day are divided into peak hours, off-peak hours, and valley hours. Through K-means clustering, an automated and refined division of charging time periods is achieved. By combining the Pearson correlation coefficient, the correlation between charging revenue and key factors such as order quantity, charging duration, and charging power is analyzed, and the influence degree of each key factor on revenue is determined. Thus, it can help the operator identify the factors that have the most significant impact on revenue, optimize the pricing strategy and resource allocation; further use the XGBoost algorithm to quantify the influence degree of each key factor on intelligent pricing, evaluate the importance of each key factor, realize the quantification of feature importance, and establish a complete quantification system for the influence of pricing factors, providing a solid data science foundation for automatically adjusting the price strategy of each charging time period in the intelligent pricing scheme of the charging station, and achieving a transformation from an experience-driven methodology to an algorithm-driven methodology. Thereby, the charging station can formulate an intelligent pricing strategy for the operation of the charging station based on the original data of users' historical charging orders, and automatically adjust the charging price of each time period according to the intelligent pricing strategy of the charging station operation, so as to guide the reasonable distribution of users' charging demands, improve the utilization efficiency of the charging equipment in the charging station, and ultimately create greater economic value for the charging station operator, achieving the balance between maximizing the revenue of the charging station and reasonable resource allocation.

[0095] In one embodiment, in step S3, based on the influence degree of the key factors output by the data mining algorithm on intelligent pricing and the original data of users' historical charging orders, a data fusion model is used to generate multiple intelligent pricing inference fusion data with inference trajectories. Specifically, it includes:

[0096] S31. In the data generation stage, optimize the chain-of-thought prompt, construct multiple questions, and use a large language model to generate multiple answers for each question. Specifically, optimizing the chain-of-thought prompt specifically includes:

[0097] Embed clear instructions for the answer generation format in the chain-of-thought prompt, adopting the structure of " <think>Inference process< / think> Inference result" to ensure the logic and coherence of the generated answers.

[0098] Fuse the degree of influence of the key factors output by the data mining algorithm on intelligent pricing with the original data of the user's historical charging orders, and inject the fused data into the chain-of-thought prompt as prior knowledge, providing rich background information and data support for the generation of the large language model.

[0099] After optimizing the chain-of-thought prompt, construct a number of (e.g., 300,000) questions, and use a large language model to generate multiple answers (e.g., 2 - 4 answers) for each question.

[0100] A large language model (LLM) is an artificial intelligence model designed to understand and generate human language. Large language models are trained on vast amounts of text data and can perform a wide range of tasks, including text summarization, translation, sentiment analysis, and more. The characteristic of large language models is their huge scale, containing billions of parameters, which helps them learn complex patterns in language data. These models are typically based on deep learning architectures such as transformers, which contribute to their impressive performance on various NLP tasks.

[0101] For example, the large language model is the claude3.7 large model. To ensure the efficiency of the generation process of the claude3.7 large model, the claude3.7 large model sets the limit of the number of tokens generated each time to 10K.

[0102] S32. In the data filtering stage, use a generative artificial intelligence model to automatically screen the multiple answers generated by the large language model for each question, and retain a number of (e.g., 200,000) intelligent pricing inference fusion data with inference trajectories after screening. Specifically, it includes:

[0103] S321. Automatically screen the multiple answers generated by the large language model for each question according to the preset rules and preset output format specifications, and screen out the answers that do not meet the preset rules and preset output format specifications.

[0104] In the first stage of data filtering, a set of preset rules are formulated based on the result criteria set by the intelligent pricing data. At the same time, according to the requirements of the preset output format specification, an automated screening mechanism is constructed based on the preset rules and the preset output format specification to automatically screen multiple answers generated by the large language model for each question, so as to accurately screen the intelligent pricing inference answers generated by the large language model, effectively eliminating the answers that do not conform to the preset rules and the preset output format specification, and improving the overall quality and reliability of the dataset for the answers that conform to the preset rules and the preset output format specification.

[0105] Among them, the preset rules include that the adjustment range of the energy fee is within the first range, the adjustment ranges of the startup fee, the detention fee, and the duration fee are within the second range, and the monthly charging income prompt is controlled within the third range. For example, if the first range is -10% to 10%, the second range is -5% to 5%, and the third range is 5% to 10%, then the preset rules include that the adjustment range of the energy fee is within -10% to 10%, the adjustment ranges of the startup fee, the detention fee, and the duration fee are within -5% to 5%, and the monthly charging income prompt is controlled within 5% to 10%.

[0106] The preset output format specification includes the output format of " <think>Inference process< / think> Inference result".

[0107] S322. Use the generative artificial intelligence model to re-evaluate the answers that do not conform to the preset rules and the preset output format specification, accurately identify the answers that are misjudged as wrong but are actually correct, and finally retain a number of (such as 200,000) intelligent pricing inference fusion data with inference tracks after screening. Specifically, it includes:

[0108] In the second stage of data filtering, for the answers determined to be non-conforming to the preset rules and the preset output format specification in the first stage, use the generative artificial intelligence model for re-evaluation, accurately identify the answers that are misjudged as wrong but are actually correct due to output format problems, and merge the answers screened by the re-evaluation with the answers screened in the first stage that conform to the preset rules and the preset output format specification to form a number of (such as 200,000) intelligent pricing inference fusion data with inference tracks finally retained after screening.

[0109] For example, the generative artificial intelligence model is the Gemin-2.0-Flash-exp model. Gemini2.0Flash (or Gemin-2.0-Flash-exp) is a lightweight and high-performance large language model (LLM) launched by Google DeepMind and is part of the Gemini 2.0 series. It is a model optimized for low-latency and high-throughput scenarios, significantly improving the response speed while maintaining strong performance, and is suitable for applications requiring real-time interaction.

[0110] In this embodiment, by combining the claude3.7 large model and the Gemin-2.0-Flash-exp model to form a data fusion model, based on the degree of influence of the key factors output by the data mining algorithm on intelligent pricing and the original data of the user's historical charging orders, in the data generation stage, the thought chain prompt words are optimized, multiple questions are constructed, and the claude3.7 large model is used to generate multiple answers for each question; in the data filtering stage, the Gemin-2.0-Flash-exp model is used to screen the multiple answers generated by the claude3.7 large model for each question, and several (such as 200,000) intelligent pricing inference fusion data with inference tracks are retained. By combining the claude3.7 large model and the Gemin-2.0-Flash-exp model to form a data fusion model, a data fusion and double-layer filtering mechanism is realized. This mechanism design of multi-source fusion and double verification significantly improves the consistency, diversity and representativeness of the training data, improves the overall quality and reliability of the data set, and provides a high-quality knowledge base for the downstream large model training.

[0111] In one embodiment, in step S4, based on the pricing fusion model and multiple intelligent pricing inference fusion data with inference tracks, an intelligent pricing strategy for charging station operation is generated. Specifically, it includes:

[0112] S41. Use the Time Series Forecasting model to obtain the key knowledge in the field of intelligent pricing. Specifically, it includes:

[0113] S411. Data preprocessing: Clean and extract features from the original data of the user's historical charging orders, extract key indicators including charging duration, charging power, order quantity, charging income, etc., and convert them into a format suitable for analysis by the sequence prediction model.

[0114] S412. Time series model selection and training: Select a sequence prediction model and model the above key indicators. By training the sequence prediction model, trends, seasonality and periodic changes in the data of the above key indicators are captured, so as to extract domain knowledge related to intelligent pricing.

[0115] S413. Domain knowledge extraction: Extract key features and rules from the trained sequence prediction model, such as the changes in charging demand during peak hours, off-peak hours and valley hours, and the charging income fluctuations during different charging time periods.

[0116] For example, the sequence prediction model is the Prophet model. The Prophet model is an open-source time series prediction model developed by Facebook (Meta), mainly used for business prediction tasks such as sales forecasting, inventory management, trend analysis, etc. The Prophet model was released in 2017 and has become widely popular due to its ease of use, robustness, and automation capabilities, making it suitable for business analysts who are not professional data scientists.

[0117] S42. Select DeepSeek R1 as the base model and use large model fine-tuning technology to fine-tune the combination of the key knowledge in the intelligent pricing field output by the sequence prediction model and the intelligent pricing inference fusion data that retains several (with inference traces) after screening, so as to generate a high-quality intelligent pricing strategy for charging station operation.

[0118] For example, the large model fine-tuning technology uses LoRA fine-tuning. LoRA (Low-Rank Adaptation) is an efficient large model fine-tuning technology proposed by Microsoft in 2021, mainly used to reduce the computational cost and memory requirements during the fine-tuning of large language models (LLMs).

[0119] Select DeepSeek-R1 as the base model. The DeepSeek-R1 model adopts a Mixture of Experts (MoE) architecture, which improves the model performance by replacing the Feed-Forward Network in the traditional Transformer with multiple expert modules. Each expert module is essentially an independent neural network. Selecting DeepSeek-R1 as the base model can significantly improve the computational efficiency and significantly reduce resource consumption while maintaining strong inference capabilities.

[0120] Normalize the previously generated intelligent pricing inference fusion data that retains several (e.g., 200,000) with inference traces after screening, and convert it into a standard dialogue format that meets the requirements of LoRA fine-tuning. Select DeepSeek R1 as the base model and use LoRA fine-tuning to fine-tune the combination of the key knowledge in the intelligent pricing field output by the sequence prediction model and the intelligent pricing inference fusion data that retains several with inference traces after screening, so as to generate a high-quality intelligent pricing strategy for charging station operation. In this way, the DeepSeek R1 large model can better understand the dynamic changes in the intelligent pricing field and generate more inference answers that meet the actual needs. The fine-tuning process uses the lora fine-tuning method for collaborative training and optimization to ensure that the performance of the DeepSeek R1 large model reaches the optimal level, so as to generate a high-quality intelligent pricing strategy for charging station operation.

[0121] In this embodiment, the sequence prediction model (such as the Prophet model) and the DeepSeek R1 large model are jointly trained as an overall pricing fusion model. Through this joint training method, the sequence prediction model and the DeepSeek R1 large model can cooperate with each other. The sequence prediction model provides accurate prediction information in the field of intelligent pricing for the DeepSeek R1 large model, while the DeepSeek R1 large model utilizes this information and performs collaborative training and optimization using LoRA fine-tuning to generate higher-quality answers, and forms an intelligent pricing strategy for the operation of charging stations from these high-quality answers. During the joint training process, the key knowledge in the field of intelligent pricing of the sequence prediction model is incorporated as prior knowledge into the LoRA fine-tuning process of the DeepSeek-R1 model, forming a fine-tuning optimization strategy for temporal knowledge fusion, realizing the organic combination of temporal insight and the capabilities of the large model. Through an end-to-end joint optimization method, the model's dynamic understanding ability of intelligent pricing is improved, and the performance of the pricing fusion model, including prediction accuracy, reasoning logic, and the practicality of answers, is continuously evaluated and optimized. According to the evaluation results, the pricing fusion model is further optimized, such as adjusting the model structure, improving the training strategy, etc., to ensure that the pricing fusion model can stably output high-quality intelligent pricing solutions.

[0122] The present invention provides a method for realizing intelligent pricing for the operation of charging stations based on data mining and fusing large model fine-tuning algorithms, which is different from the existing charging station operation pricing methods in that:

[0123] For existing large-scale charging stations, the charging price pricing strategy is generally determined based on traditional time-of-use electricity prices. This charging price pricing strategy does not fully consider the key factors affecting charging pricing in the charging station, as well as the grid load pressure of the charging station and the overall economic benefits of the charging station. The resulting unreasonable charging price structure and charging price pricing strategy are likely to cause concentrated charging behaviors among users and unreasonable distribution of charging demands, which not only increases the grid load pressure but also reduces the overall economic benefits of the charging station.

[0124] In some conventional practices of formulating charging price pricing strategies for other charging stations, although it may also involve combining the original data of users' historical charging orders with a large model to generate a charging station operation pricing strategy, the original data of users' historical charging orders is only used as auxiliary training data for the large model, and its combination with the large model cannot determine the key factors affecting charging pricing in the charging station based on the original data of users' historical charging orders, automatically adjust the price strategy for each charging period, thereby guiding the reasonable distribution of charging demands, improving the utilization efficiency of charging equipment, and ultimately achieving the purpose of creating greater economic value for the station operator.

[0125] In the technical solution of the present invention, first, the original data of the user's historical charging orders is clustered according to the charging time using the K-means clustering algorithm, realizing the automatic and fine division of charging periods. By combining the Pearson correlation coefficient, the correlation between charging income and key factors such as the number of orders, charging duration, and charging power is analyzed, and the influence degree of each key factor on income is determined. This can help operators identify the factors that have the most significant impact on income, optimize pricing strategies and resource allocation. Further, the XGBoost algorithm is used to quantify the influence degree of each key factor on intelligent pricing, evaluate the importance of each key factor, realize the quantification of feature importance, and establish a complete quantification system for the influence of pricing factors. By combining the claude3.7 large model and the Gemin-2.0-Flash-exp model to form a data fusion model, based on the influence degree of the key factors output by the data mining algorithm on intelligent pricing and the original data of the user's historical charging orders, several intelligent pricing inference fusion data with inference tracks are generated, realizing data fusion and a double-layer filtering mechanism. This mechanism design of multi-source fusion and double verification significantly improves the consistency, diversity, and representativeness of training data, and enhances the overall quality and reliability of the data set. By jointly training the sequence prediction model and the DeepSeek R1 large model as an overall pricing fusion model, integrating the key knowledge in the field of intelligent pricing of the sequence prediction model as prior knowledge into the LoRA fine-tuning process of the DeepSeek-R1 model, forming a fine-tuning optimization strategy for temporal knowledge fusion, realizing the organic combination of temporal insight and the capabilities of the large model, and improving the model's dynamic understanding ability of intelligent pricing through an end-to-end joint optimization method, and continuously evaluating and optimizing the performance of the pricing fusion model to ensure that the pricing fusion model can stably output high-quality intelligent pricing solutions. Thus, the key factors affecting charging pricing in the charging station can be determined based on the original data of the user's historical charging orders, and the price strategy for each charging period can be automatically adjusted, thereby guiding the reasonable distribution of charging demand, improving the utilization efficiency of charging equipment, and ultimately creating greater economic value for the station operator.

[0126] Based on the same concept, the present invention also provides an operation server, as Figure 3 shown. The operation server 10 includes: a memory 12, a processor 11, and one or more computer programs stored in the memory 12 and executable on the processor 11. The memory 12 and the processor 11 are coupled together through a bus system 13. When the one or more computer programs are executed by the processor 11, the following steps of a method for intelligent pricing of charging station operation provided by the embodiments of the present invention are implemented:

[0127] S1. Obtain the original data of the user's historical charging orders for the charging station;

[0128] S2. Use data mining algorithms to process the original data of users' historical charging orders, extract the key factors affecting pricing, and determine the influence degree of the key factors on intelligent pricing;

[0129] S3. Based on the influence degree of the key factors output by the data mining algorithm on intelligent pricing and the original data of users' historical charging orders, use a data fusion model to generate multiple intelligent pricing inference fusion data with inference trajectories;

[0130] S4. Based on the pricing fusion model and multiple intelligent pricing inference fusion data with inference trajectories, generate an intelligent pricing strategy for the operation of charging stations.

[0131] The method disclosed in the embodiments of the present invention described above can be applied to or implemented by the processor 11. The processor 11 may be an integrated circuit chip with signal processing capabilities. During implementation, the steps of the above method can be completed by the integrated logic circuit in the hardware of the processor 11 or instructions in the form of software. The processor 11 may be a general-purpose processor, a DSP, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 11 can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor may be a microprocessor or any conventional processor, etc. Combining the steps of the method disclosed in the embodiments of the present invention, it can be directly embodied as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium, and this storage medium is located in the memory 12. The processor 11 reads the information in the memory 12 and combines its hardware to complete the steps of the foregoing method.

[0132] It can be understood that the memory 12 in the embodiments of the present invention can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), a ferromagnetic random access memory (FRAM), a flash memory or other memory technologies, a compact disk read-only memory (CD-ROM), a digital versatile disk (DVD) or other optical disk storage, a magnetic cassette, a magnetic tape, a magnetic disk storage or other magnetic storage devices; the volatile memory can be a random access memory (RAM). By way of example but not limitation, many forms of RAM are available, such as a static random access memory (SRAM), a synchronous static random access memory (SSRAM), a dynamic random access memory (DRAM), a synchronous dynamic random access memory (SDRAM), a double data rate synchronous dynamic random access memory (DDR SDRAM), an enhanced synchronous dynamic random access memory (ESDRAM), a sync link dynamic random access memory (SLDRAM), a direct rambus random access memory (DRRAM).The memories described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memories.

[0133] It should be noted that the above embodiments of the operation server and the method embodiments belong to the same concept. For the specific implementation process, please refer to the method embodiments, and the technical features in the method embodiments are correspondingly applicable to the operation server embodiments, which will not be elaborated here.

[0134] In addition, in an exemplary embodiment, the embodiments of the present invention also provide a computer storage medium, specifically a computer-readable storage medium. For example, it includes a memory 12 that stores a computer program. One or more programs of a method for intelligent pricing of charging station operations are stored on the computer storage medium. When the one or more programs of the method for intelligent pricing of charging station operations are executed by a processor 11, the following steps of a method for intelligent pricing of charging station operations provided by the embodiments of the present invention are implemented:

[0135] S1. Obtain the original data of the user's historical charging orders of the charging station;

[0136] S2. Use a data mining algorithm to process the original data of the user's historical charging orders, extract the key factors affecting pricing, and determine the influence degree of the key factors on intelligent pricing;

[0137] S3. Based on the influence degree of the key factors output by the data mining algorithm on intelligent pricing and the original data of the user's historical charging orders, use a data fusion model to generate multiple intelligent pricing inference fusion data with inference trajectories;

[0138] S4. Based on the pricing fusion model and multiple intelligent pricing inference fusion data with inference trajectories, generate an intelligent pricing strategy for charging station operations.

[0139] It should be noted that the method program embodiments of the intelligent pricing of charging station operations on the above computer-readable storage medium and the method embodiments belong to the same concept. For the specific implementation process, please refer to the method embodiments, and the technical features in the method embodiments are correspondingly applicable to the embodiments of the above computer-readable storage medium, which will not be elaborated here.

[0140] It should be noted that in this article, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article or device including that element.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; under the idea of the present invention, the technical features in the above embodiments or different embodiments can also be combined, and the steps can be implemented in any order, and there are many other variations in different aspects of the present invention as described above. For the sake of brevity, they are not provided in detail; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for intelligent pricing of charging station operation, applied to an operation server, characterized in that The method for intelligent pricing of charging station operation includes: Obtain the original data of the user's historical charging orders at the charging station; Use data mining algorithms to process the original data of the user's historical charging orders, extract the key factors affecting pricing, and determine the influence degree of the key factors on intelligent pricing; Based on the influence degree of the key factors on intelligent pricing and the original data of the user's historical charging orders, use a data fusion model to generate multiple intelligent pricing inference fusion data with inference tracks; Based on the pricing fusion model and the multiple intelligent pricing inference fusion data with inference tracks, generate an intelligent pricing strategy for charging station operation.

2. The method for intelligent pricing of charging station operation according to claim 1, characterized in that, The use of data mining algorithms to process the original data of the user's historical charging orders, extract the key factors affecting pricing, and determine the influence degree of the key factors on intelligent pricing includes: Cluster the original data of the user's historical charging orders according to the charging time using the K-means clustering algorithm, and divide different charging periods of a day into peak periods, flat peak periods, and off-peak periods; Use the Pearson correlation coefficient to analyze the correlation between charging income and key factors including the number of orders, charging duration, and charging power, and determine the influence degree of each key factor on income; Use the XGBoost algorithm to quantify the influence degree of each key factor on intelligent pricing.

3. The method for intelligent pricing of charging station operation according to claim 2, wherein, The clustering of the original data of the user's historical charging orders according to the charging time using the K-means clustering algorithm, and dividing different charging periods of a day into peak periods, flat peak periods, and off-peak periods includes: Extract each charging period of the charging time in hours from the original data of the user's historical charging orders, and initially divide different charging periods into peak periods, flat peak periods, and off-peak periods; Extract the features related to the charging period from the original data of the user's historical charging orders: charging start time, charging end time, charging power, and charging duration; Use the K-means clustering algorithm to cluster the extracted features related to the charging period, and finally divide different charging periods of a day into peak periods, flat peak periods, and off-peak periods.

4. The method for intelligent pricing of charging station operation according to claim 1, wherein The generation of multiple intelligent pricing inference fusion data with inference tracks based on the influence degree of the key factors on intelligent pricing and the original data of the user's historical charging orders includes: In the data generation stage, optimize the chain-of-thought prompt, construct multiple questions, and use a large language model to generate multiple answers for each question; In the data filtering stage, use a generative artificial intelligence model to automatically screen the multiple answers generated by the large language model for each question, and retain several intelligent pricing inference fusion data with inference tracks after screening.

5. The method for intelligent pricing of charging station operation according to claim 4, characterized in that The optimization of the chain-of-thought prompt includes: Embed clear answer generation format instructions in the chain-of-thought prompt; Fuse the influence degree of the key factors on intelligent pricing and the original data of the user's historical charging orders, and inject the fused data into the chain-of-thought prompt as prior knowledge.

6. The method for intelligent pricing of charging station operation according to claim 4, characterized in that, In the data filtering stage, a generative artificial intelligence model is used to automatically screen multiple answers generated by the large language model for each question, and several intelligent pricing inference fusion data with inference tracks are retained after screening, including: Automatically screen multiple answers generated by the large language model for each question according to preset rules and preset output format specifications, and screen out the answers that do not meet the preset rules and preset output format specifications; Use the generative artificial intelligence model to re-evaluate the answers that do not meet the preset rules and preset output format specifications, accurately identify the answers that are misjudged as wrong but are actually correct, and finally retain several intelligent pricing inference fusion data with inference tracks after screening.

7. The method for intelligent pricing of charging station operations according to claim 6, characterized in that, The preset rules include that the adjustment range of the energy cost is within the first range, the adjustment ranges of the start-up cost, the detention cost, and the duration cost are within the second range, and the monthly charging income prompt is controlled within the third range; the preset output format specification includes <think>Inference process< / think> The output format of the reasoning result.

8. The method for intelligent pricing of charging station operation according to claim 4, characterized in that Based on the pricing fusion model and the multiple intelligent pricing inference fusion data with inference tracks, generating an intelligent pricing strategy for charging station operation includes: Use a sequence prediction model to obtain key knowledge in the field of intelligent pricing; Select DeepSeek R1 as the base model, and use the large model fine-tuning technology to fine-tune the combination of the key knowledge in the field of intelligent pricing output by the sequence prediction model and several intelligent pricing inference fusion data with inference tracks retained after screening, and generate a high-quality intelligent pricing strategy for charging station operation.

9. An operating server, characterized in that, Including: A memory, a processor, and a computer program stored on the memory and running on the processor, where when the computer program is executed by the processor, it implements the method for intelligent pricing of charging station operation according to any one of claims 1 to 8.

10. A storage medium, characterized in that, A program for the method of intelligent pricing of charging station operation is stored on the storage medium, and when the program for the method of intelligent pricing of charging station operation is executed by the processor, it implements the method for intelligent pricing of charging station operation according to any one of claims 1 to 8.

Images