Zone area charging load regulation and control system

Abstract

The embodiment of the invention relates to a zone area charging load regulation and control system. The system comprises a charging regulation and control chain, a zone area fusion terminal and a plurality of zone area charging piles. The transformer area fusion terminal in the system is connected with the charging regulation and control chain and the transformer area charging piles and is further connected with the remote power grid server side, and the power grid server side outside the system is further connected with the power grid data chain. According to the invention, the basic power utilization guarantee capability of the power distribution area can be improved, and the orderly management strength of the power distribution area on the charging pile network can also be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a charging load control system for a substation. Background Art

[0002] If a photovoltaic storage and charging integrated power station (referred to as photovoltaic storage charging station) and an electric vehicle (referred to as electric vehicle) charging pile are built in the distribution station area, the power load in the current area will superimpose the energy storage charging load and the electric vehicle charging load on the basic power load, and the power supply in the current area will also superimpose the photovoltaic power generation power and the energy storage discharge power on the basis of the conventional single power grid distribution power. Although this distribution station area with a photovoltaic storage charging station can alleviate the pressure of electric vehicle charging during peak hours through photovoltaic power generation and / or energy storage battery discharge, with the rapid growth in the number of electric vehicles, even with the supplementary power supply of the photovoltaic storage charging station, there may still be insufficient power supply in the basic power consumption field during the peak charging period. To solve this contradiction in electricity consumption, it is necessary to power regulate the charging piles, and how to power regulate the charging piles is a technical problem that the present invention needs to solve. Summary of the invention

[0003] The purpose of the present invention is to provide a substation charging load control system in view of the defects of the prior art, which system includes: a charging control chain, a substation fusion terminal and a plurality of substation charging piles. The substation fusion terminal in the system is respectively connected to the charging control chain and each substation charging pile, and is also connected to the remote power grid service end, and the power grid service end outside the system is also connected to the power grid data chain. The power grid data chain outside the system is a data storage chain based on blockchain technology, which is composed of a power grid collection subchain and a power grid control subchain. The power grid collection subchain is used to store multiple types of power supply / power consumption curves (substation power grid distribution curves, substation photovoltaic power generation curves, substation energy storage charging and discharging curves, substation basic power load curves, and substation total charging load curves) that actually occurred in each distribution substation in each historical period. The power grid control subchain is used to store the substation power grid distribution control curve set by the power grid service end at an intermediate time point in each time period for each distribution substation in the next time period. The charging regulation chain within the system is also a data storage chain based on blockchain technology, which is used to store the charging control curves of the charging piles in each area within the current distribution station area at different time periods. The substation fusion terminal in the system is used to obtain the four types of historical curves (photovoltaic power generation historical curve, energy storage charging and discharging historical curve, basic power load historical curve and total charging load historical curve) of the current distribution substation in the previous period and the grid distribution control curve of the next period through the remote grid service end at an intermediate time point in each period, and based on four types of artificial intelligence (AI) models (photovoltaic power generation prediction model, energy storage charging and discharging prediction model, basic power load prediction model and charging load prediction model), predict the four types of prediction curves (photovoltaic power generation prediction curve, energy storage charging and discharging prediction curve, basic power load prediction curve, total charging load prediction curve) of the next period according to the four types of historical curves, and determine the charging available power curve of the next period based on the grid distribution control curve and the four types of prediction curves, and set the charging control curve of all substation charging piles in the current distribution substation in the next period according to the total charging load prediction curve and the charging available power curve, and store the setting results in the charging regulation chain while sending all charging control curves to the corresponding substation charging piles. Each charging pile in the system is used to save the received charging control curve, and each time a charging vehicle is connected, the charging process of the current charging vehicle is controlled by the charging control curve corresponding to the current time period. It can be seen that the essence of the technical solution of the present invention is to predict the overall load capacity (i.e., the charging available power curve) that can be used for charging electric vehicles in the future time period of the substation with the help of the prediction function of the AI ​​model, and to subdivide and regulate the charging capacity of each charging pile in the future time period based on the predicted overall load capacity on the premise of meeting all other power loads in the future time period of the substation.The present invention can not only improve the basic power supply guarantee capability of the distribution station area, but also improve the orderly management of the charging pile network in the distribution station area.

[0004] To achieve the above-mentioned purpose, an embodiment of the present invention provides a charging load control system for a substation, the system comprising: a charging control chain, a substation fusion terminal and a plurality of substation charging piles;

[0005] The area fusion terminal is respectively connected to the charging regulation chain and each of the area charging piles; the area fusion terminal is also connected to a remote power grid service end; the power grid service end is connected to a power grid data link;

[0006] The charging control chain is a data storage chain based on blockchain technology; the charging control chain is used to store the charging control curve Pc of each charging pile in the current distribution area in each time period. x,i,j , x is the charging pile index, i is the station area index, j is the time period index; the charging control curve Pc x,i,j is a power-time data sequence; the power-time data sequence is composed of a plurality of power-time data, each of which is composed of a power information and a time point information;

[0007] The substation fusion terminal is used to confirm the time period index corresponding to the current time at an intermediate time point of each time period based on the preset mapping relationship between the time period index and time to obtain the corresponding current time period index; and use the previous and next time period indexes of the current time period index as the corresponding previous time period index and next time period index; and use the substation index information stored locally in the terminal as the corresponding current substation index; and send a first query request carrying the previous time period index, the next time period index and the current substation index to the power grid service end; and receive the power grid distribution control curve P sent back by the power grid service end net , Photovoltaic power generation history curve P pv , Energy storage charge and discharge history curve P bat , Basic power load history curve P basic And the total charging load history curve P ev Based on the preset photovoltaic power generation prediction model, energy storage charging and discharging prediction model, basic power load prediction model and charging load prediction model according to the photovoltaic power generation history curve P pv , the energy storage charge and discharge history curve P bat , the basic power load history curve P basic and the total charging load history curve P ev Predict the four types of power curves in the next period to obtain the corresponding photovoltaic power generation prediction curve Energy storage charge and discharge prediction curve Basic power load forecast curve And the total charging load prediction curve And according to the estimated relationship The charging available power curve △P of the next period is obtained by the calculation method, α is a preset coefficient, 0<α<1; and the total number of charging piles in the said area is recorded as the corresponding total number of charging piles M; and according to the said total charging load prediction curve The charging control curves of all the charging piles in the current distribution area in the next period are set by using the charging available power curve △P to obtain the corresponding M charging pile control curves And each of the electric pile control curves As a corresponding charging control curve Pc x,i＝当前台区索引,j＝下一时段索引 Store the charging control chain; and store each of the charging control curves Pc obtained this time x,i,j Send to the corresponding charging pile in the area;

[0008] Each of the charging piles in the area is used to receive the charging control curve Pc x,i,j Save it;

[0009] Each of the charging piles in the area is also used to confirm the time period index corresponding to the current time based on the mapping relationship between the time period index and time to obtain the corresponding current time period index when each charging vehicle is connected; and the charging control curve Pc matching the time period index j with the current time period index in the multiple control curves stored locally in the charging pile x,i,j as the corresponding current charging control curve; and using the current charging control curve as a guide to perform process power control on the charging process of the current charging vehicle.

[0010] Preferably, the power grid data chain is a data evidence chain based on blockchain technology; the power grid data chain is used to store two subchains, namely, the power grid collection subchain and the power grid regulation subchain; the power grid collection subchain is used to store multiple types of power supply / power consumption curves that actually occurred in each distribution station area in each historical period, and the multiple types of power supply / power consumption curves at least include the area power grid distribution curve Ps a,i,j , photovoltaic power generation curve Ps b,i,j , energy storage charging and discharging curve Ps c,i,j , Substation basic power load curve Ps d,i,j The total charging load curve Ps of the area e,i,j The power grid regulation subchain is used to store the power grid distribution control curve Pc set by the power grid service end at an intermediate time point in each jth time period for the power grid distribution operation of each distribution station in the future j+1th time period. a,i,k＝j+1, k is the time period index; the power distribution curve Ps of the substation area a,i,j 、The photovoltaic power generation curve Ps of the station area b,i,j 、The energy storage charging and discharging curve Ps of the station area c,i,j , the basic power load curve Ps of the substation area d,i,j 、The total charging load curve Ps of the station area e,i,j and the distribution control curve Pc of the power grid in the substation area a,i,k Each is a power-time data sequence; the power distribution control curve Pc a,i,k The power information corresponding to each of the time point information is a maximum power distribution power;

[0011] The power grid service end is used to extract the corresponding previous time period index, the next time period index and the current area index from the current request when receiving the first query request sent by the area fusion terminal; and the area photovoltaic power generation curve Ps whose area index i in the power grid acquisition subchain matches the current area index and the time period index j matches the previous time period index b,i,j 、The energy storage charging and discharging curve Ps of the station area c,i,j , the basic power load curve Ps of the substation area d,i,j And the total charging load curve Ps of the station area e,i,j The photovoltaic power generation history curve P is read out as the corresponding pv , the energy storage charge and discharge history curve P bat , the basic power load history curve P basic and the total charging load history curve P ev ; and the substation index i in the power grid regulation subchain matches the current substation index and the time period index k matches the next time period index of the substation power distribution control curve Pc a,i,k The corresponding power distribution control curve P is read out. net ; And the obtained power distribution control curve P net , the photovoltaic power generation history curve P pv , the energy storage charge and discharge history curve P bat , the basic power load history curve P basic and the total charging load history curve P ev Send back to the station area fusion terminal.

[0012] Preferably, the photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model, and the charging load prediction model are four AI prediction models; the photovoltaic power generation prediction model is used to calculate the photovoltaic power generation history curve P of the j-1th period input by the model. pvThe photovoltaic power generation curve of the j+1th period is predicted to obtain the corresponding photovoltaic power generation prediction curve The energy storage charge and discharge prediction model is used to calculate the energy storage charge and discharge history curve P of the j-1th period according to the input of the model. bat Predict the energy storage charge and discharge curve of the j+1th period to obtain the corresponding energy storage charge and discharge prediction curve The basic power load prediction model is used to predict the basic power load history curve P of the j-1th period according to the input of the model. basic The basic power load curve of the j+1th period is predicted to obtain the corresponding basic power load prediction curve The basic power load prediction model is used to calculate the total charging load history curve P of the j-1th period according to the input of the model. ev Predict the total charging load curve of the j+1th period to obtain the corresponding total charging load prediction curve

[0013] The photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model and the charging load prediction model are each implemented based on a type of deep learning model; the deep learning model includes at least an LSTM model, a Bi-LSTM model, a Transformer model and an Autoformer model;

[0014] When the photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model or the charging load prediction model are implemented based on the LSTM model or the Bi-LSTM model, the model structure is formed by sequentially connecting the first embedding coding module, the feature extraction network, the linear layer network and the first prediction network; the feature extraction network is implemented based on the CNN network or the MLP network; the linear layer network is formed by sequentially connecting one or more layers of fully connected layers; the first prediction network is implemented based on the LSTM model or the Bi-LSTM model; wherein the first embedding coding module is used to input the photovoltaic power generation history curve P of the model pv , the energy storage charge and discharge history curve P bat , the basic power load history curve P basic Or the total charging load history curve P ev Each of the power-time data of the corresponding power-time data sequence is subjected to power embedding coding and time embedding coding to obtain a corresponding power code x p and time code x tThe corresponding power-time coding vector v is formed, and all the obtained power-time coding vectors v are sequentially sorted to form a corresponding coding vector sequence V, which is sent to the feature extraction network; the feature extraction network is used to perform feature extraction processing on the coding vector sequence V to obtain a corresponding feature tensor X and send it to the linear layer network; the linear layer network is used to perform feature dimension reduction processing on the feature tensor X to obtain a corresponding feature vector Z and send it to the first prediction network; the first prediction network is used to perform regression prediction based on the feature vector Z to obtain the corresponding photovoltaic power generation prediction curve The energy storage charge and discharge prediction curve The basic power load forecast curve Or the total charging load prediction curve

[0015] When the photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model or the charging load prediction model are implemented based on the Transformer model or the Autoformer model, the model structure is formed by sequentially connecting the second embedded coding module and the second prediction network; the second prediction network is implemented based on the Transformer model or the Autoformer model; wherein the second embedded coding module is used to input the photovoltaic power generation history curve P pv , the energy storage charge and discharge history curve P bat , the basic power load history curve P basic Or the total charging load history curve P ev Each of the power-time data of the corresponding power-time data sequence is subjected to power embedding coding and time embedding coding to obtain a corresponding power code x p and time code x t The corresponding power-time coding vector v is formed, and all the obtained power-time coding vectors v are sequentially sorted to form a corresponding coding vector sequence V and sent to the second prediction network; the second prediction network is used to perform encoding and decoding processing according to the coding vector sequence V to obtain the corresponding photovoltaic power generation prediction curve The energy storage charge and discharge prediction curve The basic power load forecast curve Or the total charging load prediction curve

[0016] Preferably, before using the photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model and the charging load prediction model, it is necessary to prepare four corresponding model training data sets for the photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model and the charging load prediction model in advance, recorded as the corresponding first, second, third and fourth data sets; and based on the first data set, the photovoltaic power generation prediction model is model trained; and based on the second data set, the energy storage charging and discharging prediction model is model trained; and based on the third data set, the basic power load prediction model is model trained; and based on the fourth data set, the charging load prediction model is model trained; and after the model training of these four types of models is completed, the trained photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model and the charging load prediction model are installed and set on the substation fusion terminal.

[0017] Furthermore, four corresponding model training data sets are prepared in advance for the photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model and the charging load prediction model, and are recorded as the corresponding first, second, third and fourth data sets. Specifically, four long time series equal-length curves with the same start and end time and the same time sampling interval are constructed by performing big data collection on the photovoltaic power generation curve, energy storage charging and discharging curve, basic power load curve and total charging load curve of the distribution station area where the substation fusion terminal is located, and are recorded as the corresponding first, second, third and fourth curves; and the first, second, third and fourth curves are sequentially sliced ​​based on the preset time period length to obtain the corresponding first slice sequence S 1 , the second slice sequence S 2 , the third slice sequence S 3 and the fourth slice sequence S 4 ; and based on the first slice sequence S 1 The second slice sequence S 2 , the third slice sequence S 3 and the fourth slice sequence S 4 Construct four model training data sets and record them as the corresponding first, second, third and fourth data sets;

[0018] Among them, the first slice sequence S 1 By multiple first slice curves s 1,r The second slice sequence S 2 By multiple second slice curves s 2,r The third slice sequence S 3 By multiple third slice curves s3,r The fourth slice sequence S 4 By multiple fourth slice curves s 4,r composition;

[0019] The first data set includes R-2 first data records; the first data record includes a first training curve and a first label curve; the first training curve and the first label curve of the first first data record are respectively a first slicing curve s 1,r＝1 and the first slice curve s 1,r＝3 The first training curve and the first label curve of the second first data record are respectively the first slicing curve s 1,r＝2 and the first slice curve s 1,r＝4 ; and so on, the first training curve and the first label curve of the R-2th first data record are respectively the first slicing curve s 1,r＝R-2 and the first slice curve s 1,r＝R ;

[0020] The second data set includes R-2 second data records; the second data record includes a second training curve and a second label curve; the second training curve and the second label curve of the first second data record are respectively a second slicing curve s 2,r＝1 and the second slice curve s 2,r＝3 The second training curve and the second label curve of the second second data record are respectively the second slicing curve s 2,r＝2 and the second slice curve s 2,r＝4 ; and so on, the second training curve and the second label curve of the R-2th second data record are the second slicing curves s 2,r＝R-2 and the second slice curve s 2,r＝R ;

[0021] The third data set includes R-2 third data records; the third data record includes a third training curve and a third label curve; the third training curve and the third label curve of the first third data record are respectively a third slicing curve s 3,r＝1 and the third slice curve s 3,r＝3 ; The third training curve and the third label curve of the second third data record are respectively the third slicing curve s 3,r＝2 and the third slice curve s 3,r＝4 ; and so on, the third training curve and the third label curve of the R-2th third data record are respectively the third slicing curve s 3,r＝R-2 and the third slice curve s 3,r＝R ;

[0022] The fourth data set includes R-2 fourth data records; the fourth data record includes a fourth training curve and a fourth label curve; the fourth training curve and the fourth label curve of the first fourth data record are respectively a fourth slicing curve s 4,r＝1 and the fourth slice curve s 4,r＝3 ; The fourth training curve and the fourth label curve of the second fourth data record are respectively the fourth slicing curve s 4,r＝2 and the fourth slice curve s 4,r＝4 ; and so on, the fourth training curve and the fourth label curve of the R-2th fourth data record are respectively the fourth slicing curve s 4,r＝R-2 and the fourth slice curve s 4,r＝R .

[0023] Furthermore, the performing model training on the photovoltaic power generation prediction model based on the first data set specifically includes:

[0024] Step 61, based on a preset first segmentation ratio, the first data set is divided into two sub-data sets which are recorded as a corresponding first training set and a first evaluation set;

[0025] Wherein, both the first training set and the first evaluation set are composed of a plurality of the first data records; the ratio of the total number of records in the first training set and the first evaluation set satisfies the first segmentation ratio;

[0026] Step 62, taking the first first data record of the first training set as the corresponding current training record;

[0027] Step 63, inputting the first training curve of the current training record into the photovoltaic power generation prediction model to perform photovoltaic power generation curve prediction processing to obtain a corresponding first prediction curve; and bringing the first prediction curve and the first label curve of the current training record into a preset first model loss function; and based on a preset first model optimizer, performing a round of modulation on the model parameters of the photovoltaic power generation prediction model in a direction in which the first model loss function reaches a minimum value;

[0028] Wherein, the first model loss function includes at least an L1 loss function, an L2 loss function and a SmoothL1 loss function; the first model optimizer includes at least an SGD optimizer and an Adam optimizer;

[0029] Step 64, identifying whether the current training record is the last first data record of the first training set; if so, going to step 65; if not, taking the next first data record of the first training set as the new current training record and returning to step 63;

[0030] Step 65, perform a round of traversal on all the first data records of the first evaluation set; and in this round of traversal, use the first data record currently traversed as the corresponding current evaluation record; and input the first training curve of the current evaluation record into the photovoltaic power generation prediction model to perform photovoltaic power generation curve prediction processing to obtain the corresponding second prediction curve; and form a corresponding first prediction-label pair by the second prediction curve and the first label curve of the current evaluation record; and at the end of this round of traversal, bring all the first prediction-label pairs obtained into the preset first model evaluation function to calculate and obtain the corresponding first evaluation value;

[0031] Wherein, the first model evaluation function at least includes an RMSE function;

[0032] Step 66, identifying whether the first evaluation value meets the preset first evaluation value range; if not, returning to step 62 to continue training; if so, stopping training and confirming that the model training of the photovoltaic power generation prediction model is completed.

[0033] Furthermore, the performing model training on the energy storage charge and discharge prediction model based on the second data set specifically includes:

[0034] Step 71, based on a preset second segmentation ratio, the second data set is divided into two sub-data sets recorded as corresponding second training sets and second evaluation sets;

[0035] Wherein, both the second training set and the second evaluation set are composed of a plurality of the second data records; the ratio of the total number of records in the second training set and the second evaluation set satisfies the second segmentation ratio;

[0036] Step 72, taking the first second data record of the second training set as the corresponding current training record;

[0037] Step 73, inputting the second training curve of the current training record into the energy storage charge and discharge prediction model to perform energy storage charge and discharge curve prediction processing to obtain a corresponding third prediction curve; and bringing the third prediction curve and the second label curve of the current training record into a preset second model loss function; and based on a preset second model optimizer, performing a round of modulation on the model parameters of the energy storage charge and discharge prediction model in a direction to minimize the second model loss function;

[0038] Wherein, the second model loss function includes at least an L1 loss function, an L2 loss function and a SmoothL1 loss function; the second model optimizer includes at least an SGD optimizer and an Adam optimizer;

[0039] Step 74, identifying whether the current training record is the last second data record of the second training set; if so, going to step 75; if not, taking the next second data record of the second training set as the new current training record and returning to step 73;

[0040] Step 75, perform a round of traversal on all the second data records of the second evaluation set; and in this round of traversal, use the second data record currently traversed as the corresponding current evaluation record; and input the second training curve of the current evaluation record into the energy storage charge and discharge prediction model to perform energy storage charge and discharge curve prediction processing to obtain a corresponding fourth prediction curve; and form a corresponding second prediction-label pair from the fourth prediction curve and the second label curve of the current evaluation record; and at the end of this round of traversal, bring all the obtained second prediction-label pairs into the preset second model evaluation function to calculate and obtain the corresponding second evaluation value;

[0041] Wherein, the second model evaluation function at least includes an RMSE function;

[0042] Step 76, identifying whether the second evaluation value meets the preset second evaluation value range; if not, returning to step 72 to continue training; if so, stopping training and confirming that the model training of the energy storage charge and discharge prediction model is completed.

[0043] Furthermore, the performing model training on the basic power load prediction model based on the third data set specifically includes:

[0044] Step 81, based on a preset third segmentation ratio, the third data set is divided into two sub-data sets which are recorded as a corresponding third training set and a third evaluation set;

[0045] Wherein, the third training set and the third evaluation set are both composed of a plurality of the third data records; the ratio of the total number of records of the third training set and the third evaluation set meets the third segmentation ratio;

[0046] Step 82, taking the first third data record of the third training set as the corresponding current training record;

[0047] Step 83, inputting the third training curve of the current training record into the basic power load forecasting model to perform basic power load curve forecasting processing to obtain a corresponding fifth prediction curve; and bringing the fifth prediction curve and the third label curve of the current training record into a preset third model loss function; and based on a preset third model optimizer, performing a round of modulation on the model parameters of the basic power load forecasting model in a direction to minimize the third model loss function;

[0048] Wherein, the third model loss function includes at least an L1 loss function, an L2 loss function and a SmoothL1 loss function; the third model optimizer includes at least an SGD optimizer and an Adam optimizer;

[0049] Step 84, identifying whether the current training record is the last third data record of the third training set; if so, going to step 85; if not, taking the next third data record of the third training set as the new current training record and returning to step 83;

[0050] Step 85, perform a round of traversal on all the third data records of the third evaluation set; and in this round of traversal, use the third data record currently traversed as the corresponding current evaluation record; and input the third training curve of the current evaluation record into the basic power load prediction model to perform energy storage charge and discharge curve prediction processing to obtain the corresponding sixth prediction curve; and form a corresponding third prediction-label pair from the sixth prediction curve and the third label curve of the current evaluation record; and at the end of this round of traversal, bring all the obtained third prediction-label pairs into the preset third model evaluation function to calculate and obtain the corresponding third evaluation value;

[0051] Wherein, the third model evaluation function at least includes RMSE function;

[0052] Step 86, identifying whether the third evaluation value meets the preset third evaluation value range; if not, returning to step 82 to continue training; if so, stopping training and confirming that the model training of the basic power load prediction model is completed.

[0053] Furthermore, the performing model training on the charging load prediction model based on the fourth data set specifically includes:

[0054] Step 91, based on a preset fourth segmentation ratio, the fourth data set is divided into two sub-data sets which are recorded as a corresponding fourth training set and a fourth evaluation set;

[0055] Wherein, the fourth training set and the fourth evaluation set are both composed of a plurality of the fourth data records; the ratio of the total number of records of the fourth training set and the fourth evaluation set meets the fourth segmentation ratio;

[0056] Step 92, taking the first fourth data record of the fourth training set as the corresponding current training record;

[0057] Step 93, inputting the fourth training curve of the current training record into the charging load prediction model to perform charging load curve prediction processing to obtain the corresponding seventh prediction curve; and bringing the seventh prediction curve and the fourth label curve of the current training record into a preset fourth model loss function; and based on the preset fourth model optimizer, performing a round of modulation on the model parameters of the charging load prediction model in the direction of minimizing the fourth model loss function;

[0058] Wherein, the fourth model loss function includes at least an L1 loss function, an L2 loss function and a SmoothL1 loss function; the fourth model optimizer includes at least an SGD optimizer and an Adam optimizer;

[0059] Step 94, identifying whether the current training record is the last fourth data record of the fourth training set; if so, going to step 95; if not, taking the next fourth data record of the fourth training set as the new current training record and returning to step 93;

[0060] Step 95, perform a round of traversal on all the fourth data records of the fourth evaluation set; and in this round of traversal, use the fourth data record currently traversed as the corresponding current evaluation record; and input the fourth training curve of the current evaluation record into the charging load prediction model to perform charging load curve prediction processing to obtain the corresponding eighth prediction curve; and form a corresponding fourth prediction-label pair by the eighth prediction curve and the fourth label curve of the current evaluation record; and at the end of this round of traversal, bring all the fourth prediction-label pairs obtained into the preset fourth model evaluation function to calculate and obtain the corresponding fourth evaluation value;

[0061] Wherein, the fourth model evaluation function at least includes RMSE function;

[0062] Step 96, identifying whether the fourth evaluation value meets the preset fourth evaluation value range; if not, returning to step 92 to continue training; if so, stopping training and confirming that the model training of the charging load prediction model is completed.

[0063] Preferably, the station area fusion terminal is specifically used for The charging control curves of all the charging piles in the current distribution area in the next period are set by using the charging available power curve △P to obtain the corresponding M charging pile control curves hour:

[0064] The rated charging power of each charging pile in the station area is recorded as the corresponding rated charging power And for the M rated charging power Perform the sum calculation to get the corresponding rated power sum

[0065] The total charging load prediction curve Divide the time period into N time periods on average to obtain a plurality of corresponding segmented curves, where N is a preset positive integer; perform mean square error calculation on each segmented curve to obtain a corresponding segmented mean square error; calculate the absolute difference between each two segmented mean square errors to obtain a corresponding pairwise absolute difference; count the number of pairwise absolute differences exceeding a preset absolute difference threshold to obtain a corresponding first number; identify whether the first number is greater than 1; if not, set the corresponding control mode to a static mode; if yes, set the corresponding control mode to a dynamic mode; the control mode includes a static mode and a dynamic mode;

[0066] and identifying the obtained regulation mode;

[0067] If the control mode is the static mode, the minimum power on the charging available power curve ΔP is used as the corresponding upper limit of the total available power p max ; and based on the total available power upper limit p max , the total rated power and all the rated charging powers stated Set the corresponding M electric pile control curves

[0068] If the control mode is the dynamic mode, based on the charging available power curve ΔP and all the rated charging powers Set the corresponding M electric pile control curves

[0069] Further, the station area fusion terminal is specifically used for the available total power upper limit p max , the total rated power and all the rated charging powers stated Set the corresponding M electric pile control curves When each of the electric pile control curves The setting method is:

[0070]

[0071] Wherein, t is the control curve of the electric pile The time point information, is the electric pile control curve The power information corresponding to each of the time point information t.

[0072] Further, the station area fusion terminal is specifically used for the charging available power curve △P based on all the rated charging power Set the corresponding M electric pile control curves When each of the electric pile control curves The setting method is:

[0073]

[0074] Wherein, t is the control curve of the electric pile The time point information, is the electric pile control curve The power information corresponding to each of the time point information t.

[0075] Preferably, the charging pile in the area is specifically used for performing process power control on the charging process of the current charging vehicle under the guidance of the current charging control curve:

[0076] Step 131, taking the access time of the current charging vehicle as the corresponding current start time;

[0077] Step 132, extracting a control curve after the current start time on the current charging control curve as the corresponding current charging curve; and continuously charging the currently charged vehicle based on the current charging curve;

[0078] Step 133, during the continuous charging process based on the current charging curve, real-time monitoring is performed on whether the charge state of the currently charged vehicle has reached the target specified by the owner of the currently charged vehicle, and real-time monitoring is performed on whether the current time has reached the end time of the current charging curve; and when the charge state of the currently charged vehicle reaches the target specified by the owner, the process goes to step 134; and when the current time has reached the end time of the current charging curve, the time period index corresponding to the current time is confirmed again based on the mapping relationship between the time period index and time to obtain a new current time period index, and the charging control curve Pc whose time period index j matches the new current time period index among the multiple control curves stored locally in the electric pile is matched with the charging control curve Pc x,i,j As the new current charging control curve, and return to step 132 to enter the next continuous charging process;

[0079] Step 134, stopping charging of the currently charging vehicle.

[0080] The embodiment of the present invention provides a substation charging load control system, which includes: a charging control chain, a substation fusion terminal and a plurality of substation charging piles. The substation fusion terminal in the system is respectively connected to the charging control chain and each substation charging pile, and is also connected to the remote power grid service end, and the power grid service end outside the system is also connected to the power grid data chain. The power grid data chain outside the system is a data storage chain based on blockchain technology, which is composed of a power grid collection subchain and a power grid control subchain. The power grid collection subchain is used to store multiple types of power supply / power consumption curves (substation power grid distribution curve, substation photovoltaic power generation curve, substation energy storage charging and discharging curve, substation basic power load curve, substation charging total load curve) that actually occurred in each distribution substation in each historical period, and the power grid control subchain is used to store the substation power grid distribution control curve set by the power grid service end at an intermediate time point in each time period for each distribution substation in the next time period. The charging control chain in the system is also a data storage chain based on blockchain technology, which is used to store the charging control curves of each substation charging pile in each time period in the current distribution substation. The substation fusion terminal in the system is used to obtain the four types of historical curves (photovoltaic power generation historical curve, energy storage charging and discharging historical curve, basic power load historical curve and total charging load historical curve) of the current distribution substation in the previous period and the grid distribution control curve of the next period through the remote power grid service end at an intermediate time point in each period, and based on four types of AI models (photovoltaic power generation prediction model, energy storage charging and discharging prediction model, basic power load prediction model and charging load prediction model), predict the four types of prediction curves (photovoltaic power generation prediction curve, energy storage charging and discharging prediction curve, basic power load prediction curve, total charging load prediction curve) of the next period according to the four types of historical curves, and determine the charging available power curve of the next period based on the grid distribution control curve of the next period and the four types of prediction curves, and set the charging control curve of all substation charging piles in the current distribution substation in the next period according to the total charging load prediction curve and the charging available power curve, and store the setting results in the charging regulation chain while sending all charging control curves to their corresponding substation charging piles. Each charging pile in the system is used to save the received charging control curve, and each time a charging vehicle is connected, the charging process of the current charging vehicle is guided by the charging control curve corresponding to the current time period. It can be seen that the essence of the technical solution of the embodiment of the present invention is to predict the overall load capacity (i.e., the charging available power curve) that can be used for charging electric vehicles in the future time period of the substation with the help of the prediction function of the AI ​​model, and to subdivide and regulate the charging capacity of each charging pile in the future time period based on the predicted overall load capacity on the premise of meeting all other power loads in the future time period of the substation. The embodiment of the present invention not only improves the basic power supply guarantee capacity of the distribution station area, but also improves the orderly management of the charging pile network by the distribution station area. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] Figure 1 A schematic diagram of the structure of a charging load control system for a substation provided by an embodiment of the present invention;

[0082] Figure 2 Schematic diagram of two types of model structures provided for embodiments of the present invention. DETAILED DESCRIPTION

[0083] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0084] The charging load control system 1 provided in the embodiment of the present invention is as follows: Figure 1 As shown in the structural diagram of a substation charging load control system provided by an embodiment of the present invention, the system mainly includes: a charging control chain 11, a substation fusion terminal 12 and multiple substation charging piles 13. Among them, the substation fusion terminal 12 is respectively connected to the charging control chain 11 and each substation charging pile 13; the substation fusion terminal 12 is also connected to the remote power grid service terminal 3; the power grid service terminal 3 is also connected to the power grid data link 2.

[0085] Before describing the functions of the various system components of the charging load control system 1 , the power grid data link 2 and the power grid service end 3 of the embodiment of the present invention are briefly described below.

[0086] The power grid data chain 2 of the embodiment of the present invention is a data evidence chain based on blockchain technology; the power grid data chain 2 is used to store two sub-chains, namely, the power grid collection sub-chain and the power grid regulation sub-chain:

[0087] 1) The power grid collection subchain is used to store multiple types of power supply / power consumption curves that actually occurred in each distribution station area in each historical period; among them, the multiple types of power supply / power consumption curves at least include the area power grid distribution curve Ps a,i,j , photovoltaic power generation curve Ps b,i,j , energy storage charging and discharging curve Ps c,i,j , Substation basic power load curve Ps d,i,j The total charging load curve Ps of the area e,i,j ; i is the station area index, j is the time period index;

[0088] 2) The power grid regulation subchain is used to store the power grid distribution control curve Pc set by the power grid service end 3 at an intermediate time point in each jth time period for the power grid distribution operation of each distribution station in the future j+1th time period a,i,k＝j+1 , k is the time period index.

[0089] It should be noted that the above-mentioned power distribution curve Ps a,i,j , photovoltaic power generation curve Ps b,i,j , energy storage charging and discharging curve Ps c,i,j , Substation basic power load curve Ps d,i,j , total charging load curve Ps e,i,j And the distribution control curve Pc of the power grid in the area a,i,k Each is a power-time data sequence. It should also be noted that the power-time data sequence of the embodiment of the present invention is composed of multiple power-time data, and each power-time data is composed of a power information and a time point information; that is to say, the various types of power curves mentioned in the embodiment of the present invention are actually a sampled data sequence composed of discretely sampled power-time data. It should also be noted that the above-mentioned power distribution control curve Pc a,i,k The power information corresponding to each time point is a maximum distribution power or a maximum available distribution power. The distribution control curve Pc of the power grid in this area a,i,k In general, it is used to control the transformers that connect each distribution substation to the main power grid (hereinafter referred to as the substation-main network access transformer).

[0090] The power grid service end 3 of the embodiment of the present invention is the service end of the superior management organization of all power distribution substations, which can be a service interface, server, system, platform or cloud. The functions of the power grid service end 3 of the embodiment of the present invention are relatively complex. Because the technical solution of the embodiment of the present invention is mainly based on the substation charging load control system 1, all functions of the power grid service end 3 are not fully described, but two service functions of the power grid service end 3 related to the embodiment of the present invention need to be explained.

[0091] One of the service functions related to the embodiment of the present invention is: the power grid service end 3 is used to predict various types of power loads of each distribution sub-area in the future j+1 period based on a series of AI models installed on the service end at an intermediate time point in each j-th period according to multiple types of power supply / power consumption curves of all distribution sub-areas in the j-1 period on the power grid acquisition sub-chain, and further plan the power grid distribution behavior of the sub-area-main network access transformer of each distribution sub-area in the future j+1 period according to the overall power distribution capacity of the power grid side and the prediction results of various types of power consumption loads of all distribution sub-areas, and output the corresponding sub-area power grid distribution control curve Pc a,i,kIt should be noted that the transformers connected to the main network of each distribution area will also use the distribution area power control curve Pc corresponding to the distribution area where it is located and the current period on the power grid control subchain in each period. a,i,k To guide the corresponding power grid distribution operations to the distribution station area where it is located.

[0092] Another service function related to the embodiment of the present invention is: the power grid service end 3 is also used to extract the corresponding previous time period index, next time period index and current area index from the current request when receiving the first query request sent by the area fusion terminal 12; and match the area index i in the power grid acquisition subchain with the current area index and the time period index j with the previous time period index. b,i,j , energy storage charging and discharging curve Ps c,i,j , Substation basic power load curve Ps d,i,j The total charging load curve Ps of the area e,i,j The corresponding photovoltaic power generation history curve P is read out. pv , Energy storage charge and discharge history curve P bat , Basic power load history curve P basic And the total charging load history curve P ev ; and the power distribution control curve Pc of the power grid in the power grid regulation subchain is matched with the current power grid index and the time period index k is matched with the next time period index a,i,k Read out as the corresponding power distribution control curve P net ; And the obtained power distribution control curve P net , Photovoltaic power generation history curve P pv , Energy storage charge and discharge history curve P bat , Basic power load history curve P basic And the total charging load history curve P ev Send back to the station area fusion terminal 12.

[0093] After briefly describing the power grid data link 2 and the power grid service end 3 above, the functions of each system component of the substation charging load control system 1 according to the embodiment of the present invention are described below.

[0094] (I) Charging control chain 11:

[0095] The charging control chain 11 of the embodiment of the present invention is a data storage chain based on blockchain technology; the charging control chain 11 is used to store the charging control curve Pc of each charging pile 13 in each time period in the current distribution station area. x,i,j ; Where x is the charging pile index, i is the station area index, and j is the time period index; each charging control curve Pc x,i,jIt is a power-time data sequence. As can be seen from the foregoing, the power-time data sequence of the embodiment of the present invention is composed of multiple power-time data, and each power-time data is composed of a power information and a time point information. It should be noted that each charging control curve Pc x,i,j The power information corresponding to each time point can be understood as the maximum available charging power at a time point; in principle, if each charging control curve Pc x,i,j The configuration is appropriate and each charging pile 13 in each area is based on the corresponding charging control curve Pc x,i,j By controlling the charging process itself, the probability of occurrences such as excessive charging load on a single pile or excessive overall charging load can be effectively reduced.

[0096] (II) Area fusion terminal 12:

[0097] The substation fusion terminal 12 of the embodiment of the present invention is used to confirm the time period index corresponding to the current time at an intermediate time point of each time period based on the mapping relationship between the preset time period index and time to obtain the corresponding current time period index; and use the previous and next time period indexes of the current time period index as the corresponding previous time period index and next time period index; and use the substation index information stored locally in the terminal as the corresponding current substation index; and send the first query request carrying the previous time period index, the next time period index and the current substation index to the power grid service end 3; and receive the power grid distribution control curve P sent back by the power grid service end 3 net , Photovoltaic power generation history curve P pv , Energy storage charge and discharge history curve P bat , Basic power load history curve P basic And the total charging load history curve P ev Based on the preset photovoltaic power generation prediction model, energy storage charging and discharging prediction model, basic power load prediction model and charging load prediction model, the photovoltaic power generation history curve P pv , Energy storage charge and discharge history curve P bat , Basic power load history curve P basic And the total charging load history curve P ev Predict the four types of power curves in the next period to obtain the corresponding photovoltaic power generation prediction curve Energy storage charge and discharge prediction curve Basic power load forecast curve And the total charging load prediction curve And according to the estimated relationship The calculation method is used to obtain the charging available power curve △P for the next period; and the total number of charging piles 13 in the area is recorded as the corresponding total number of charging piles M; and according to the total charging load prediction curve The charging control curves of all charging piles 13 in the current distribution area are set with the charging available power curve △P to obtain the corresponding M charging pile control curves in the next period And each charging pile control curve As a corresponding charging control curve Pc x,i＝当前台区索引,j＝下一时段索引 Store the charging control chain 11; and store the various charging control curves Pc obtained this time x,i,j Send to the corresponding charging station 13.

[0098] The mapping relationship between the time period index and time is a preset time-time period index correspondence rule, which can be customized based on the actual application needs. Under normal circumstances, the corresponding rule stipulates that the time period index is an integer starting from 1, and the time period length is set to a fixed value L, and a start time (the time format of the start time should at least include information such as year, month, day, hour, minute and second) is specified as the starting zero point t 0 , and stipulate that the time period index corresponding to the first time period after the start time is 1, the time period index corresponding to the second time period after the first time period is 2, and so on. The time period index is increased by 1 every time period length L in the time series, and the mapping function F(t now )=Ce ili ng((t now -t 0 ) / L), where t now is the current time, Ceiling() is the rounding up function.

[0099] The above estimated relationship α in the figure is a preset coefficient, and the value range of the preset coefficient is: 0<α<1; the setting of the preset coefficient can be configured based on application requirements. Under normal circumstances, the preset coefficient α will correspond to the transformer capacity conversion coefficient of the substation-main network access transformer of the distribution substation where the current substation fusion terminal 12 is located, for example 80%.

[0100] The above-mentioned photovoltaic power generation prediction model, energy storage charging and discharging prediction model, basic power load prediction model, and charging load prediction model are four types of AI prediction models installed and configured on the substation fusion terminal 12; the respective uses of these four types of AI prediction models are:

[0101] 1) The photovoltaic power generation prediction model is used to calculate the photovoltaic power generation history curve P of the j-1 period according to the input of the model pv Predict the photovoltaic power generation curve in the j+1th period to obtain the corresponding photovoltaic power generation prediction curve

[0102] 2) The energy storage charge and discharge prediction model is used to calculate the energy storage charge and discharge history curve P of the j-1 period according to the input of the model bat Predict the energy storage charge and discharge curve in the j+1th period to obtain the corresponding energy storage charge and discharge prediction curve

[0103] 3) The basic power load forecasting model is used to calculate the basic power load history curve P of the j-1 period according to the input of the model basic The basic electricity load curve of the j+1 period is predicted to obtain the corresponding basic electricity load prediction curve

[0104] 4) The basic power load forecasting model is used to calculate the total charging load history curve P of the j-1th period according to the input of the model ev Predict the total charging load curve of the j+1th period to obtain the corresponding total charging load prediction curve

[0105] It should be noted that the above-mentioned photovoltaic power generation prediction model, energy storage charging and discharging prediction model, basic power load prediction model and charging load prediction model are each implemented based on a type of deep learning model; the deep learning model mentioned in the embodiment of the present invention includes at least LSTM model, Bi-LSTM model, Transformer model and Autoformer model. If it is implemented based on the LSTM model or the Bi-LSTM model, the constructed model structure is similar and can be recorded as the first type of model structure; if it is implemented based on the Transformer model or the Autoformer model, the constructed model structure is also similar and can be recorded as the second type of model structure; but there are differences between the first and second type of model structures. The following is combined with Figure 2 The following is a schematic diagram of two types of model structures provided in an embodiment of the present invention, which briefly describes the two types of model structures.

[0106] When the photovoltaic power generation prediction model, energy storage charging and discharging prediction model, basic power load prediction model or charging load prediction model is implemented based on the LSTM model or Bi-LSTM model, the corresponding model structure is the first type of model structure. Figure 2 As shown, the first type of model structure is specifically composed of a first embedded coding module, a feature extraction network, a linear layer network and a first prediction network connected in sequence; wherein, the feature extraction network is implemented based on a CNN network or an MLP network; the linear layer network is composed of one or more layers of fully connected layers connected in sequence; and the first prediction network is implemented based on an LSTM model or a Bi-LSTM model.

[0107] The component functions of each model component of the first type of model structure are:

[0108] 1) The first embedded coding module is used to input the photovoltaic power generation history curve P of the model pv , Energy storage charge and discharge history curve P bat , Basic power load history curve P basic Or charging total load history curve P ev Each power-time data of the corresponding power-time data sequence is subjected to power embedding coding and time embedding coding to obtain the corresponding power code x p and time code x t The corresponding power-time coding vector v is formed, and all the obtained power-time coding vectors v are sequentially sorted to form a corresponding coding vector sequence V and sent to the feature extraction network;

[0109] 2) The feature extraction network is used to perform feature extraction processing on the encoding vector sequence V to obtain the corresponding feature tensor X and send it to the linear layer network;

[0110] 3) The linear layer network is used to perform feature dimension reduction processing on the feature tensor X to obtain the corresponding feature vector Z and send it to the first prediction network;

[0111] 4) The first prediction network is used to perform regression prediction based on the feature vector Z to obtain the corresponding photovoltaic power generation prediction curve Energy storage charge and discharge prediction curve Basic power load forecast curve Or charging total load prediction curve

[0112] When the photovoltaic power generation prediction model, energy storage charging and discharging prediction model, basic power load prediction model or charging load prediction model is implemented based on the Transformer model or the Autoformer model, the corresponding model structure is the second type of model structure. Figure 2 As shown, the second type of model structure is formed by sequentially connecting a second embedding coding module and a second prediction network; wherein the second prediction network is implemented based on a Transformer model or an Autoformer model.

[0113] The component functions of each model component of the second type of model structure are:

[0114] 1) The second embedded coding module is used to input the photovoltaic power generation history curve P of the model pv , Energy storage charge and discharge history curve P bat , Basic power load history curve P basic Or charging total load history curve P ev Each power-time data of the corresponding power-time data sequence is subjected to power embedding coding and time embedding coding to obtain the corresponding power code x p and time code x tThe corresponding power-time coding vector v is formed, and all the obtained power-time coding vectors v are sequentially sorted to form a corresponding coding vector sequence V and sent to the second prediction network;

[0115] 2) The second prediction network is used to encode and decode the coding vector sequence V to obtain the corresponding photovoltaic power generation prediction curve Energy storage charge and discharge prediction curve Basic power load forecast curve Or charging total load prediction curve

[0116] It should be noted that before using the photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model and the charging load prediction model, four corresponding model training data sets should be prepared in advance for the photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model and the charging load prediction model, and recorded as the corresponding first, second, third and fourth data sets; and the photovoltaic power generation prediction model is trained based on the first data set; and the energy storage charging and discharging prediction model is trained based on the second data set; and the basic power load prediction model is trained based on the third data set; and the charging load prediction model is trained based on the fourth data set; and after the model training of these four types of models is completed, the trained photovoltaic power generation prediction model, energy storage charging and discharging prediction model, basic power load prediction model and charging load prediction model are installed and set on the substation fusion terminal 12.

[0117] Among them, the above four corresponding model training data sets are prepared in advance for the photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model and the charging load prediction model, which are recorded as the corresponding first, second, third and fourth data sets, specifically including:

[0118] Step A1, by collecting big data of the photovoltaic power generation curve, energy storage charging and discharging curve, basic power load curve and total charging load curve of the distribution station area where the station fusion terminal 12 is located, four long time series equal length curves with the same start and end time and the same time sampling interval are constructed and recorded as the corresponding first, second, third and fourth curves;

[0119] Here, when performing big data collection, the embodiment of the present invention should at least collect data for multiple months in units of months, so that at least the model can learn the power supply / power consumption change rules in the off-season / peak season within a year. If the multi-month collection amount exceeds one year, it can also help the model learn the power supply / power consumption change rules brought about by the change of four seasons. It is better if multi-year data collection can be performed in units of years, which can not only help the model learn the power supply / power consumption change rules in different years, but also help the model learn the power supply / power consumption change rules year by year.

[0120] Step A2: sequentially slice the first, second, third and fourth curves based on the preset time period to obtain a corresponding first slice sequence S. 1 , the second slice sequence S 2 , the third slice sequence S 3 and the fourth slice sequence S 4 ;

[0121] The time period length mentioned here is the time period length mentioned in the mapping relationship between time period index and time in the previous text; the first slice sequence S 1 By multiple first slice curves s 1,r The total number of slices R is a positive integer. Since the lengths of the first, second, third, and fourth curves are the same by default, the total number of slices R is a positive integer calculated by rounding down the total length of any curve divided by the time period length. The second slice sequence S 2 By multiple second slice curves s 2,r Composition: The third slice sequence S 3 By multiple third slice curves s 3,r Composition: The fourth slice sequence S 4 By multiple fourth slice curves s 4,r composition;

[0122] Step A3, based on the first slice sequence S 1 , the second slice sequence S 2 , the third slice sequence S 3 and the fourth slice sequence S 4 Four model training data sets are constructed and recorded as the corresponding first, second, third and fourth data sets.

[0123] Here, as shown above, the prediction method of the four types of AI models in the embodiment of the present invention is to predict the next time period information of the current time period based on the previous time period information of the current time period, so when constructing the first, second, third and fourth data sets, it is defaulted to construct the corresponding training curve and label curve from the first, second, third and fourth slice sequences by collecting the interval slice curves. Therefore, it can be known that the total number of records of the first, second, third and fourth data sets is R-2; specifically as shown below:

[0124] 1) The first data set includes R-2 first data records; the first data record includes a first training curve and a first label curve; the first training curve and the first label curve of the first first data record are respectively the first slicing curve s 1,r＝1 and the first slice curve s 1,r＝3 ; The first training curve and the first label curve of the second first data record are respectively the first slicing curve s 1,r＝2and the first slice curve s 1,r＝4 ; By analogy, the first training curve and the first label curve of the R-2th first data record are the first slicing curves s 1,r＝R-2 and the first slice curve s 1,r＝R ;

[0125] 2) The second data set includes R-2 second data records; the second data record includes a second training curve and a second label curve; the second training curve and the second label curve of the first second data record are respectively the second slicing curve s 2,r＝1 and the second slice curve s 2,r＝3 ; The second training curve and the second label curve of the second second data record are respectively the second slicing curve s 2,r＝2 and the second slice curve s 2,r＝4 ; By analogy, the second training curve and the second label curve of the R-2 second data record are the second slicing curves s 2,r＝R-2 and the second slice curve s 2,r＝R ;

[0126] 3) The third data set includes R-2 third data records; the third data record includes a third training curve and a third label curve; the third training curve and the third label curve of the first third data record are respectively the third slicing curve s 3,r＝1 and the third slice curve s 3,r＝3 ; The third training curve and the third label curve of the second third data record are respectively the third slicing curve s 3,r＝2 and the third slice curve s 3,r＝4 ; By analogy, the third training curve and the third label curve of the R-2 third data record are the third slicing curve s 3,r＝R-2 and the third slice curve s 3,r＝R ;

[0127] 4) The fourth data set includes R-2 fourth data records; the fourth data record includes a fourth training curve and a fourth label curve; the fourth training curve and the fourth label curve of the first fourth data record are respectively the fourth slicing curve s 4,r＝1 and the fourth slice curve s 4,r＝3 ; The fourth training curve and the fourth label curve of the second fourth data record are respectively the fourth slicing curve s 4,r＝2 and the fourth slice curve s 4,r＝4 ; By analogy, the fourth training curve and the fourth label curve of the R-2 fourth data record are the fourth slicing curve s 4,r＝R-2 and the fourth slice curve s 4,r＝R .

[0128] In addition, the aforementioned photovoltaic power generation prediction model is trained based on the first data set, specifically including:

[0129] Step B1, based on a preset first segmentation ratio, split the first data set into two sub-data sets, which are recorded as a corresponding first training set and a first evaluation set;

[0130] The first segmentation ratio is a preset ratio parameter, such as 8:2, 9:1, etc.; the first training set and the first evaluation set are both composed of a plurality of first data records; the ratio of the total number of records in the first training set and the first evaluation set satisfies the first segmentation ratio;

[0131] Step B2, taking the first first data record of the first training set as the corresponding current training record;

[0132] Step B3, inputting the first training curve of the current training record into the photovoltaic power generation prediction model to perform photovoltaic power generation curve prediction processing to obtain the corresponding first prediction curve; and bringing the first prediction curve and the first label curve of the current training record into the preset first model loss function; and based on the preset first model optimizer, performing a round of modulation on the model parameters of the photovoltaic power generation prediction model in the direction of minimizing the first model loss function;

[0133] The first model loss function includes at least an L1 loss function, an L2 loss function, and a SmoothL1 loss function; the first model optimizer includes at least an SGD optimizer and an Adam optimizer;

[0134] Step B4, identifying whether the current training record is the last first data record of the first training set; if so, going to step B5; if not, taking the next first data record of the first training set as the new current training record and returning to step B3;

[0135] Step B5, perform a round of traversal on all first data records of the first evaluation set; and in this round of traversal, use the currently traversed first data record as the corresponding current evaluation record; and input the first training curve of the current evaluation record into the photovoltaic power generation prediction model to perform photovoltaic power generation curve prediction processing to obtain the corresponding second prediction curve; and form a corresponding first prediction-label pair by the second prediction curve and the first label curve of the current evaluation record; and at the end of this round of traversal, bring all the obtained first prediction-label pairs into the preset first model evaluation function to calculate and obtain the corresponding first evaluation value;

[0136] Wherein, the first model evaluation function at least includes the RMSE function;

[0137] Step B6, identifying whether the first evaluation value meets the preset first evaluation value range; if not, returning to step B2 to continue training; if so, stopping training and confirming that the model training of the photovoltaic power generation prediction model is completed.

[0138] Here, the first evaluation value range is a preset value range.

[0139] In addition, the aforementioned model training of the energy storage charging and discharging prediction model based on the second data set specifically includes:

[0140] Step C1, based on a preset second segmentation ratio, split the second data set into two sub-data sets, which are recorded as a corresponding second training set and a second evaluation set;

[0141] The second segmentation ratio is a preset ratio parameter, such as 8:2, 9:1, etc.; the second training set and the second evaluation set are both composed of a plurality of second data records; the ratio of the total number of records in the second training set to the second evaluation set satisfies the second segmentation ratio;

[0142] Step C2, taking the first second data record of the second training set as the corresponding current training record;

[0143] Step C3, inputting the second training curve of the current training record into the energy storage charge and discharge prediction model to perform energy storage charge and discharge curve prediction processing to obtain a corresponding third prediction curve; and bringing the third prediction curve and the second label curve of the current training record into a preset second model loss function; and based on the preset second model optimizer, performing a round of modulation on the model parameters of the energy storage charge and discharge prediction model in a direction to minimize the second model loss function;

[0144] The second model loss function includes at least an L1 loss function, an L2 loss function, and a SmoothL1 loss function; the second model optimizer includes at least an SGD optimizer and an Adam optimizer;

[0145] Step C4, identifying whether the current training record is the last second data record of the second training set; if so, going to step C5; if not, taking the next second data record of the second training set as the new current training record and returning to step C3;

[0146] Step C5, perform a round of traversal on all second data records of the second evaluation set; and in this round of traversal, use the currently traversed second data record as the corresponding current evaluation record; and input the second training curve of the current evaluation record into the energy storage charge and discharge prediction model to perform energy storage charge and discharge curve prediction processing to obtain the corresponding fourth prediction curve; and form a corresponding second prediction-label pair by the fourth prediction curve and the second label curve of the current evaluation record; and at the end of this round of traversal, bring all the obtained second prediction-label pairs into the preset second model evaluation function to calculate and obtain the corresponding second evaluation value;

[0147] Wherein, the second model evaluation function at least includes the RMSE function;

[0148] Step C6, identifying whether the second evaluation value meets the preset second evaluation value range; if not, returning to step C2 to continue training; if so, stopping training and confirming that the model training of the energy storage charge and discharge prediction model is completed.

[0149] Here, the second evaluation value range is a preset value range.

[0150] In addition, the aforementioned model training of the basic power load prediction model based on the third data set specifically includes:

[0151] Step D1, based on a preset third segmentation ratio, split the third data set into two sub-data sets recorded as a corresponding third training set and a third evaluation set;

[0152] The third segmentation ratio is a preset ratio parameter, such as 8:2, 9:1, etc.; the third training set and the third evaluation set are both composed of a plurality of third data records; the ratio of the total number of records in the third training set and the third evaluation set satisfies the third segmentation ratio;

[0153] Step D2, taking the first third data record of the third training set as the corresponding current training record;

[0154] Step D3, inputting the third training curve of the current training record into the basic power load forecasting model to perform basic power load curve forecasting processing to obtain the corresponding fifth prediction curve; and bringing the fifth prediction curve and the third label curve of the current training record into the preset third model loss function; and based on the preset third model optimizer, performing a round of modulation on the model parameters of the basic power load forecasting model in the direction of minimizing the third model loss function;

[0155] The third model loss function includes at least an L1 loss function, an L2 loss function, and a SmoothL1 loss function; the third model optimizer includes at least an SGD optimizer and an Adam optimizer;

[0156] Step D4, identifying whether the current training record is the last third data record of the third training set; if so, going to step D5; if not, taking the next third data record of the third training set as the new current training record and returning to step D3;

[0157] Step D5, perform a round of traversal on all third data records of the third evaluation set; and in this round of traversal, use the third data record currently traversed as the corresponding current evaluation record; and input the third training curve of the current evaluation record into the basic power load prediction model to perform energy storage charging and discharging curve prediction processing to obtain the corresponding sixth prediction curve; and the sixth prediction curve and the third label curve of the current evaluation record form a corresponding third prediction-label pair; and at the end of this round of traversal, bring all the obtained third prediction-label pairs into the preset third model evaluation function to calculate and obtain the corresponding third evaluation value;

[0158] Wherein, the third model evaluation function at least includes the RMSE function;

[0159] Step D6, identifying whether the third evaluation value meets the preset third evaluation value range; if not, returning to step D2 to continue training; if so, stopping training and confirming that the model training of the basic power load prediction model is completed.

[0160] Here, the third evaluation value range is a preset numerical range.

[0161] In addition, the aforementioned model training of the charging load prediction model based on the fourth data set specifically includes:

[0162] Step E1, based on a preset fourth segmentation ratio, dividing the fourth data set into two sub-data sets, which are recorded as a corresponding fourth training set and a fourth evaluation set;

[0163] The fourth segmentation ratio is a preset ratio parameter, such as 8:2, 9:1, etc.; the fourth training set and the fourth evaluation set are both composed of a plurality of fourth data records; the ratio of the total number of records of the fourth training set and the fourth evaluation set satisfies the fourth segmentation ratio;

[0164] Step E2, taking the first fourth data record of the fourth training set as the corresponding current training record;

[0165] Step E3, inputting the fourth training curve of the current training record into the charging load prediction model to perform charging load curve prediction processing to obtain the corresponding seventh prediction curve; and bringing the seventh prediction curve and the fourth label curve of the current training record into the preset fourth model loss function; and based on the preset fourth model optimizer, performing a round of modulation on the model parameters of the charging load prediction model in the direction of minimizing the fourth model loss function;

[0166] The fourth model loss function includes at least an L1 loss function, an L2 loss function, and a SmoothL1 loss function; the fourth model optimizer includes at least an SGD optimizer and an Adam optimizer;

[0167] Step E4, identifying whether the current training record is the last fourth data record of the fourth training set; if so, going to step E5; if not, taking the next fourth data record of the fourth training set as the new current training record and returning to step E3;

[0168] Step E5, perform a round of traversal on all fourth data records of the fourth evaluation set; and in this round of traversal, use the fourth data record currently traversed as the corresponding current evaluation record; and input the fourth training curve of the current evaluation record into the charging load prediction model to perform charging load curve prediction processing to obtain the corresponding eighth prediction curve; and the eighth prediction curve and the fourth label curve of the current evaluation record form a corresponding fourth prediction-label pair; and at the end of this round of traversal, bring all the obtained fourth prediction-label pairs into the preset fourth model evaluation function to calculate and obtain the corresponding fourth evaluation value;

[0169] Wherein, the fourth model evaluation function at least includes the RMSE function;

[0170] Step E6, identifying whether the fourth evaluation value meets the preset fourth evaluation value range; if not, returning to step E2 to continue training; if so, stopping training and confirming that the model training of the charging load prediction model is completed.

[0171] Here, the fourth evaluation value range is a preset numerical range.

[0172] It should also be noted that the station area fusion terminal 12 is specifically used to calculate the total charging load prediction curve. The charging control curves of all charging piles 13 in the current distribution area are set with the charging available power curve △P to obtain the corresponding M charging pile control curves in the next period hour:

[0173] Step F1: record the rated charging power of each charging pile 13 in each area as the corresponding rated charging power And for M rated charging power Perform the sum calculation to get the corresponding rated power sum

[0174] in,

[0175] Step F2, and calculate the total charging load prediction curve Divide the time periods into N time periods on average to obtain a corresponding plurality of segmented curves, where N is a preset positive integer; calculate the mean square error of each segmented curve to obtain the corresponding segmented mean square error; calculate the absolute difference between the mean square errors of each two segmented curves to obtain the corresponding pairwise absolute difference; count the number of pairwise absolute differences exceeding the preset absolute difference threshold to obtain the corresponding first number; identify whether the first number is greater than 1; if not, set the corresponding control mode to a static mode; if so, set the corresponding control mode to a dynamic mode; the control mode includes a static mode and a dynamic mode;

[0176] Here, by comparing the absolute differences of the segmented mean square errors of the curve / signal, the smooth / violent fluctuation state of the curve / signal can be identified; the absolute difference threshold of the embodiment of the present invention is a preset threshold parameter; when the first number is greater than 1, it indicates that the current curve / signal is the total charging load prediction curve There are at least three segmented curves / signals with large fluctuations in amplitude. At this time, the embodiment of the present invention believes that the charging demand of electric vehicles in the future period may change greatly. For this reason, the control mode corresponding to the future period is set to dynamic mode. It can be seen from the following that in the dynamic mode, the charging control of the electric pile will be based on a larger available charging range. The advantage of this is that the charging efficiency of electric vehicles can be improved while ensuring the power supply capacity of other power loads. When the first number is less than or equal to 1, that is, 0 or 1, it indicates that the current curve / signal is the total charging load prediction curve The amplitude changes relatively slowly in the future period. At this time, the embodiment of the present invention believes that the charging demand of electric vehicles in the future period is stable and the time distribution is relatively balanced. Therefore, the control mode corresponding to the future period is set to the static mode. It can be seen from the following that in the static mode, the charging control of the electric pile will be performed based on a smaller maximum charging upper limit. The advantage of this is that the charging load of the electric vehicle can be balanced and regulated under the premise of ensuring the power supply capacity of other power loads;

[0177] Step F3, identifying the obtained regulation mode;

[0178] Step F4: If the control mode is the static mode, the minimum power on the charging available power curve △P is used as the corresponding upper limit of the total available power p max ; and based on the available total power upper limit p max , total rated power and all rated charging powers Set the corresponding M charging pile control curves

[0179] Here, the station area integration terminal 12 is specifically used to generate a power supply based on the upper limit p of the total available power. max , total rated power and all rated charging powers Set the corresponding M charging pile control curves When each charging pile controls the curve The setting method is:

[0180]

[0181] Among them, t is the charging pile control curve Time point information, The control curve of the electric pile The power information corresponding to each time point t above;

[0182] Step F5: If the control mode is dynamic mode, based on the charging available power curve △P and all rated charging powers Set the corresponding M charging pile control curves

[0183] Here, the station integration terminal 12 is specifically used to calculate the charging power curve ΔP based on the charging available power curve and all rated charging powers. Set the corresponding M charging pile control curves When each charging pile controls the curve The setting method is:

[0184]

[0185] Among them, t is the charging pile control curve Time point information, The control curve of the electric pile The power information corresponding to each time point t.

[0186] (III) Area charging pile 13:

[0187] The charging piles 13 in each area of ​​the embodiment of the present invention are used to receive the charging control curve Pc x,i,j to save.

[0188] Each charging pile 13 in each area is also used to confirm the time period index corresponding to the current time based on the mapping relationship between the time period index and time to obtain the corresponding current time period index when each charging vehicle is connected; and to match the charging control curve Pc with the time period index j in the multiple control curves stored locally in the charging pile and the current time period index. x,i,j as the corresponding current charging control curve; and using the current charging control curve as a guide to perform process power control on the charging process of the current charging vehicle.

[0189] It should be noted that the charging pile 13 in the area is specifically used to control the process power of the charging process of the current charging vehicle under the guidance of the current charging control curve:

[0190] Step G1, taking the access time of the current charging vehicle as the corresponding current start time;

[0191] Step G2, extracting the control curve after the current start time on the current charging control curve as the corresponding current charging curve; and continuously charging the currently charged vehicle based on the current charging curve;

[0192] Step G3, in the process of continuous charging based on the current charging curve, whether the charge state of the current charging vehicle has reached the target specified by the owner of the current charging vehicle is monitored in real time, and whether the current time has reached the end time of the current charging curve is monitored in real time; and when the charge state of the current charging vehicle reaches the target specified by the owner, go to step G4; and when the current time has reached the end time of the current charging curve, the time period index corresponding to the current time is confirmed again based on the mapping relationship between the time period index and the time to obtain a new current time period index, and the charging control curve Pc matching the time period index j in the multiple control curves stored locally in the electric pile with the new current time period index is matched. x,i,j As the new current charging control curve, and return to step G2 to enter the next continuous charging process;

[0193] Step G4, stop charging the currently charging vehicle.

[0194] In summary, an embodiment of the present invention provides a substation charging load control system, which includes: a charging control chain, a substation fusion terminal and a plurality of substation charging piles. The substation fusion terminal in the system is respectively connected to the charging control chain and each substation charging pile, and is also connected to the remote power grid service end, and the power grid service end outside the system is also connected to the power grid data chain. The power grid data chain outside the system is a data storage chain based on blockchain technology, which is composed of a power grid collection subchain and a power grid control subchain. The power grid collection subchain is used to store multiple types of power supply / power consumption curves (substation power grid distribution curves, substation photovoltaic power generation curves, substation energy storage charging and discharging curves, substation basic power load curves, and substation total charging load curves) that actually occurred in each distribution substation in each historical period. The power grid control subchain is used to store the substation power grid distribution control curve set by the power grid service end at an intermediate time point in each time period for each distribution substation in the next time period. The charging regulation chain within the system is also a data storage chain based on blockchain technology, which is used to store the charging control curves of the charging piles in each area within the current distribution station area at different time periods. The substation fusion terminal in the system is used to obtain the four types of historical curves (photovoltaic power generation historical curve, energy storage charging and discharging historical curve, basic power load historical curve and total charging load historical curve) of the current distribution substation in the previous period and the grid distribution control curve of the next period through the remote power grid service end at an intermediate time point in each period, and based on four types of AI models (photovoltaic power generation prediction model, energy storage charging and discharging prediction model, basic power load prediction model and charging load prediction model), predict the four types of prediction curves (photovoltaic power generation prediction curve, energy storage charging and discharging prediction curve, basic power load prediction curve, total charging load prediction curve) of the next period according to the four types of historical curves, and determine the charging available power curve of the next period based on the grid distribution control curve of the next period and the four types of prediction curves, and set the charging control curve of all substation charging piles in the current distribution substation in the next period according to the total charging load prediction curve and the charging available power curve, and store the setting results in the charging regulation chain while sending all charging control curves to their corresponding substation charging piles. Each charging pile in the system is used to save the received charging control curve, and each time a charging vehicle is connected, the charging process of the current charging vehicle is guided by the charging control curve corresponding to the current time period. It can be seen that the essence of the technical solution of the embodiment of the present invention is to predict the overall load capacity (i.e., the charging available power curve) that can be used for charging electric vehicles in the future time period of the substation with the help of the prediction function of the AI ​​model, and to subdivide and regulate the charging capacity of each charging pile in the future time period based on the predicted overall load capacity on the premise of meeting all other power loads in the future time period of the substation. The embodiment of the present invention not only improves the basic power supply guarantee capacity of the distribution station area, but also improves the orderly management of the charging pile network by the distribution station area.

[0195] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0196] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A charging load control system for a substation, characterized in that: The system includes: a charging regulation chain, a substation fusion terminal and a plurality of substation charging piles; The area fusion terminal is respectively connected to the charging regulation chain and each of the area charging piles; the area fusion terminal is also connected to a remote power grid service end; the power grid service end is connected to a power grid data link; The charging control chain is a data storage chain based on blockchain technology; the charging control chain is used to store the charging control curve Pc of each charging pile in the current distribution area in each time period. x,i,j , x is the charging pile index, i is the station area index, j is the time period index; the charging control curve Pc x,i,j is a power-time data sequence; the power-time data sequence is composed of a plurality of power-time data, each of which is composed of a power information and a time point information; The substation fusion terminal is used to confirm the time period index corresponding to the current time at an intermediate time point of each time period based on the preset mapping relationship between the time period index and time to obtain the corresponding current time period index; and use the previous and next time period indexes of the current time period index as the corresponding previous time period index and next time period index; and use the substation index information stored locally in the terminal as the corresponding current substation index; and send a first query request carrying the previous time period index, the next time period index and the current substation index to the power grid service end; and receive the power grid distribution control curve P sent back by the power grid service end net , Photovoltaic power generation history curve P pv , Energy storage charge and discharge history curve P bat , Basic power load history curve P basic And the total charging load history curve P ev Based on the preset photovoltaic power generation prediction model, energy storage charging and discharging prediction model, basic power load prediction model and charging load prediction model according to the photovoltaic power generation history curve P pv , the energy storage charge and discharge history curve P bat , the basic power load history curve P basic and the total charging load history curve P ev Predict the four types of power curves in the next period to obtain the corresponding photovoltaic power generation prediction curve Energy storage charge and discharge prediction curve Basic power load forecast curve And the total charging load prediction curve And according to the estimated relationship The charging available power curve △P of the next period is obtained by the calculation method, α is a preset coefficient, 0<α<1; and the total number of charging piles in the said area is recorded as the corresponding total number of charging piles M; and according to the said total charging load prediction curve The charging control curves of all the charging piles in the current distribution area in the next period are set by using the charging available power curve △P to obtain the corresponding M charging pile control curves And each of the electric pile control curves As a corresponding charging control curve Pc x,i＝当前台区索引,j＝下一时段索引 Store the charging control chain; and store each of the charging control curves Pc obtained this time x,i,j Send to the corresponding charging pile in the area; Each of the charging piles in the area is used to receive the charging control curve Pc x,i,j Save it; Each of the charging piles in the area is also used to confirm the time period index corresponding to the current time based on the mapping relationship between the time period index and time to obtain the corresponding current time period index when each charging vehicle is connected; and the charging control curve Pc matching the time period index j with the current time period index in the multiple control curves stored locally in the charging pile x,i,j as the corresponding current charging control curve; and using the current charging control curve as a guide to perform process power control on the charging process of the current charging vehicle.

2. The charging load control system of the substation according to claim 1, characterized in that: The power grid data chain is a data storage chain based on blockchain technology; the power grid data chain is used to store two subchains, namely the power grid collection subchain and the power grid regulation subchain; the power grid collection subchain is used to store multiple types of power supply / power consumption curves that actually occurred in each distribution station area in each historical period, and the multiple types of power supply / power consumption curves at least include the area power grid distribution curve Ps a,i,j , photovoltaic power generation curve Ps b,i,j , energy storage charging and discharging curve Ps c,i,j , Substation basic power load curve Ps d,i,j The total charging load curve Ps of the area e,i,j The power grid regulation subchain is used to store the power grid distribution control curve Pc set by the power grid service end at an intermediate time point in each jth time period for the power grid distribution operation of each distribution station in the future j+1th time period. a,i,k＝j+1 , k is the time period index; the power distribution curve Ps of the substation area a,i,j 、The photovoltaic power generation curve Ps of the station area b,i,j 、The energy storage charging and discharging curve Ps of the station area c,i,j , the basic power load curve Ps of the substation area d,i,j 、The total charging load curve Ps of the station area e,i,j and the distribution control curve Pc of the power grid in the substation area a,i,k Each is a power-time data sequence; the power distribution control curve Pc a,i,k The power information corresponding to each of the time point information is a maximum power distribution power; The power grid service end is used to extract the corresponding previous time period index, the next time period index and the current area index from the current request when receiving the first query request sent by the area fusion terminal; The photovoltaic power generation curve Ps of the area whose area index i in the power grid acquisition subchain matches the current area index and whose time period index j matches the previous time period index b,i,j 、The energy storage charging and discharging curve Ps of the station area c,i,j , the basic power load curve Ps of the substation area d,i,j And the total charging load curve Ps of the station area e,i,j The photovoltaic power generation history curve P is read out as the corresponding pv , the energy storage charge and discharge history curve P bat , the basic power load history curve P basic and the total charging load history curve P ev ; and the substation index i in the power grid regulation subchain matches the current substation index and the time period index k matches the next time period index of the substation power distribution control curve Pc a,i,k The corresponding power distribution control curve P is read out. net ; And the obtained power distribution control curve P net , the photovoltaic power generation history curve P pv , the energy storage charge and discharge history curve P bat , the basic power load history curve P basic and the total charging load history curve P ev Send back to the station area fusion terminal.

3. The charging load control system of the substation according to claim 1, characterized in that: The photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model, and the charging load prediction model are four AI prediction models; the photovoltaic power generation prediction model is used to calculate the photovoltaic power generation history curve P of the j-1 period input by the model. pv The photovoltaic power generation curve of the j+1th period is predicted to obtain the corresponding photovoltaic power generation prediction curve The energy storage charge and discharge prediction model is used to calculate the energy storage charge and discharge history curve P of the j-1th period according to the input of the model. bat Predict the energy storage charge and discharge curve of the j+1th period to obtain the corresponding energy storage charge and discharge prediction curve The basic power load prediction model is used to predict the basic power load history curve P of the j-1th period according to the input of the model. basic The basic power load curve of the j+1th period is predicted to obtain the corresponding basic power load prediction curve The basic power load prediction model is used to calculate the total charging load history curve P of the j-1th period according to the input of the model. ev Predict the total charging load curve of the j+1th period to obtain the corresponding total charging load prediction curve The photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model and the charging load prediction model are each implemented based on a type of deep learning model; the deep learning model includes at least an LSTM model, a Bi-LSTM model, a Transformer model and an Autoformer model; When the photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model or the charging load prediction model are implemented based on the LSTM model or the Bi-LSTM model, the model structure is formed by sequentially connecting the first embedding coding module, the feature extraction network, the linear layer network and the first prediction network; the feature extraction network is implemented based on the CNN network or the MLP network; the linear layer network is formed by sequentially connecting one or more layers of fully connected layers; the first prediction network is implemented based on the LSTM model or the Bi-LSTM model; wherein the first embedding coding module is used to input the photovoltaic power generation history curve P of the model pv , the energy storage charge and discharge history curve P bat , the basic power load history curve P basic Or the total charging load history curve P ev Each of the power-time data of the corresponding power-time data sequence is subjected to power embedding coding and time embedding coding to obtain a corresponding power code x p and time code x t The corresponding power-time coding vector v is formed, and all the obtained power-time coding vectors v are sequentially sorted to form a corresponding coding vector sequence V, which is sent to the feature extraction network; the feature extraction network is used to perform feature extraction processing on the coding vector sequence V to obtain a corresponding feature tensor X and send it to the linear layer network; the linear layer network is used to perform feature dimension reduction processing on the feature tensor X to obtain a corresponding feature vector Z and send it to the first prediction network; the first prediction network is used to perform regression prediction based on the feature vector Z to obtain the corresponding photovoltaic power generation prediction curve The energy storage charge and discharge prediction curve The basic power load forecast curve Or the total charging load prediction curve When the photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model or the charging load prediction model are implemented based on the Transformer model or the Autoformer model, the model structure is formed by sequentially connecting the second embedded coding module and the second prediction network; the second prediction network is implemented based on the Transformer model or the Autoformer model; wherein the second embedded coding module is used to input the photovoltaic power generation history curve P pv , the energy storage charge and discharge history curve P bat , the basic power load history curve P basic Or the total charging load history curve P ev Each of the power-time data of the corresponding power-time data sequence is subjected to power embedding coding and time embedding coding to obtain a corresponding power code x p and time code x t The corresponding power-time coding vector v is formed, and all the obtained power-time coding vectors v are sequentially sorted to form a corresponding coding vector sequence V and sent to the second prediction network; the second prediction network is used to perform encoding and decoding processing according to the coding vector sequence V to obtain the corresponding photovoltaic power generation prediction curve The energy storage charge and discharge prediction curve The basic power load forecast curve Or the total charging load prediction curve 4. The charging load control system of the substation according to claim 3 is characterized in that: Before using the photovoltaic power generation prediction model, the energy storage charge and discharge prediction model, the basic power load prediction model and the charging load prediction model, it is necessary to prepare four corresponding model training data sets for the photovoltaic power generation prediction model, the energy storage charge and discharge prediction model, the basic power load prediction model and the charging load prediction model in advance, recorded as the corresponding first, second, third and fourth data sets; and perform model training on the photovoltaic power generation prediction model based on the first data set; and perform model training on the energy storage charge and discharge prediction model based on the second data set; and perform model training on the basic power load prediction model based on the third data set; and perform model training on the charging load prediction model based on the fourth data set; After the model training of these four types of models is completed, the trained photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model and the charging load prediction model are installed and set on the substation fusion terminal.

5. The charging load control system of the substation according to claim 4, characterized in that: The four corresponding model training data sets prepared in advance for the photovoltaic power generation prediction model, the energy storage charging and discharging prediction model, the basic power load prediction model and the charging load prediction model are recorded as the corresponding first, second, third and fourth data sets, specifically: by collecting big data of the photovoltaic power generation curve, energy storage charging and discharging curve, basic power load curve and total charging load curve of the distribution station area where the substation fusion terminal is located, four long time series equal length curves with the same start and end time and the same time sampling interval are constructed and recorded as the corresponding first, second, third and fourth curves; and based on the preset time period length, the first, second, third and fourth curves are sequentially sliced ​​to obtain the corresponding first slice sequence S1, second slice sequence S2, third slice sequence S3 and fourth slice sequence S4; and based on the first slice sequence S1, the second slice sequence S2, the third slice sequence S3 and the fourth slice sequence S4, four model training data sets are constructed and recorded as the corresponding first, second, third and fourth data sets; The first slicing sequence S1 is composed of a plurality of first slicing curves s 1,r The second slice sequence S2 is composed of a plurality of second slice curves s 2,r The third slicing sequence S3 consists of a plurality of third slicing curves s 3,r The fourth slice sequence S4 consists of a plurality of fourth slice curves s 4,r composition; The first data set includes R-2 first data records; the first data record includes a first training curve and a first label curve; the first training curve and the first label curve of the first first data record are respectively a first slicing curve s 1,r＝1 and the first slice curve s 1,r＝3 ; The first training curve and the first label curve of the second first data record are respectively the first slicing curve s 1,r＝2 and the first slice curve s 1,r＝4 ; and so on, the first training curve and the first label curve of the R-2th first data record are respectively the first slicing curve s 1,r＝R-2 and the first slice curve s 1,r＝R ; The second data set includes R-2 second data records; the second data record includes a second training curve and a second label curve; the second training curve and the second label curve of the first second data record are respectively a second slicing curve s 2,r＝1 and the second slice curve s 2,r＝3 ; The second training curve and the second label curve of the second second data record are respectively the second slicing curve s 2,r＝2 and the second slice curve s 2,r＝4 ; and so on, the second training curve and the second label curve of the R-2th second data record are the second slicing curves s 2,r＝R-2 and the second slice curve s 2,r＝R ; The third data set includes R-2 third data records; the third data record includes a third training curve and a third label curve; the third training curve and the third label curve of the first third data record are respectively a third slicing curve s 3,r＝1 and the third slice curve s 3,r＝3 ; The third training curve and the third label curve of the second third data record are respectively the third slicing curve s 3,r＝2 and the third slice curve s 3,r＝4 ; and so on, the third training curve and the third label curve of the R-2th third data record are respectively the third slicing curve s 3,r＝R-2 and the third slice curve s 3,r＝R ; The fourth data set includes R-2 fourth data records; the fourth data record includes a fourth training curve and a fourth label curve; the fourth training curve and the fourth label curve of the first fourth data record are respectively a fourth slicing curve s 4,r＝1 and the fourth slice curve s 4,r＝3 ; The fourth training curve and the fourth label curve of the second fourth data record are respectively the fourth slicing curve s 4,r＝2 and the fourth slice curve s 4,r＝4 ; and so on, the fourth training curve and the fourth label curve of the R-2th fourth data record are respectively the fourth slicing curve s 4,r＝R-2 and the fourth slice curve s 4,r＝R .

6. The charging load control system of the substation according to claim 5, characterized in that: The performing model training on the photovoltaic power generation prediction model based on the first data set specifically includes: Step 61, based on a preset first segmentation ratio, the first data set is divided into two sub-data sets which are recorded as a corresponding first training set and a first evaluation set; Wherein, both the first training set and the first evaluation set are composed of a plurality of the first data records; the ratio of the total number of records in the first training set and the first evaluation set satisfies the first segmentation ratio; Step 62, taking the first first data record of the first training set as the corresponding current training record; Step 63, inputting the first training curve of the current training record into the photovoltaic power generation prediction model to perform photovoltaic power generation curve prediction processing to obtain a corresponding first prediction curve; and bringing the first prediction curve and the first label curve of the current training record into a preset first model loss function; and based on a preset first model optimizer, performing a round of modulation on the model parameters of the photovoltaic power generation prediction model in a direction in which the first model loss function reaches a minimum value; Wherein, the first model loss function includes at least an L1 loss function, an L2 loss function and a SmoothL1 loss function; the first model optimizer includes at least an SGD optimizer and an Adam optimizer; Step 64, identifying whether the current training record is the last first data record of the first training set; if so, going to step 65; if not, taking the next first data record of the first training set as the new current training record and returning to step 63; Step 65, perform a round of traversal on all the first data records of the first evaluation set; and in this round of traversal, use the first data record currently traversed as the corresponding current evaluation record; and input the first training curve of the current evaluation record into the photovoltaic power generation prediction model to perform photovoltaic power generation curve prediction processing to obtain the corresponding second prediction curve; and form a corresponding first prediction-label pair by the second prediction curve and the first label curve of the current evaluation record; and at the end of this round of traversal, bring all the first prediction-label pairs obtained into the preset first model evaluation function to calculate and obtain the corresponding first evaluation value; Wherein, the first model evaluation function at least includes an RMSE function; Step 66, identifying whether the first evaluation value meets the preset first evaluation value range; if not, returning to step 62 to continue training; if so, stopping training and confirming that the model training of the photovoltaic power generation prediction model is completed.

7. The charging load control system of the substation according to claim 5, characterized in that: The performing model training on the energy storage charging and discharging prediction model based on the second data set specifically includes: Step 71, based on a preset second segmentation ratio, the second data set is divided into two sub-data sets recorded as corresponding second training sets and second evaluation sets; Wherein, both the second training set and the second evaluation set are composed of a plurality of the second data records; the ratio of the total number of records in the second training set and the second evaluation set satisfies the second segmentation ratio; Step 72, taking the first second data record of the second training set as the corresponding current training record; Step 73, inputting the second training curve of the current training record into the energy storage charge and discharge prediction model to perform energy storage charge and discharge curve prediction processing to obtain a corresponding third prediction curve; and bringing the third prediction curve and the second label curve of the current training record into a preset second model loss function; and based on a preset second model optimizer, performing a round of modulation on the model parameters of the energy storage charge and discharge prediction model in a direction to minimize the second model loss function; Wherein, the second model loss function includes at least an L1 loss function, an L2 loss function and a SmoothL1 loss function; the second model optimizer includes at least an SGD optimizer and an Adam optimizer; Step 74, identifying whether the current training record is the last second data record of the second training set; if so, going to step 75; if not, taking the next second data record of the second training set as the new current training record and returning to step 73; Step 75, perform a round of traversal on all the second data records of the second evaluation set; and in this round of traversal, use the second data record currently traversed as the corresponding current evaluation record; and input the second training curve of the current evaluation record into the energy storage charge and discharge prediction model to perform energy storage charge and discharge curve prediction processing to obtain a corresponding fourth prediction curve; and form a corresponding second prediction-label pair from the fourth prediction curve and the second label curve of the current evaluation record; and at the end of this round of traversal, bring all the obtained second prediction-label pairs into the preset second model evaluation function to calculate and obtain the corresponding second evaluation value; Wherein, the second model evaluation function at least includes an RMSE function; Step 76, identifying whether the second evaluation value meets the preset second evaluation value range; if not, returning to step 72 to continue training; if so, stopping training and confirming that the model training of the energy storage charge and discharge prediction model is completed.

8. The charging load control system of the substation according to claim 5, characterized in that: The performing model training on the basic power load prediction model based on the third data set specifically includes: Step 81, based on a preset third segmentation ratio, the third data set is divided into two sub-data sets which are recorded as a corresponding third training set and a third evaluation set; Wherein, the third training set and the third evaluation set are both composed of a plurality of the third data records; the ratio of the total number of records of the third training set and the third evaluation set satisfies the third segmentation ratio; Step 82, taking the first third data record of the third training set as the corresponding current training record; Step 83, inputting the third training curve of the current training record into the basic power load forecasting model to perform basic power load curve forecasting processing to obtain a corresponding fifth prediction curve; and bringing the fifth prediction curve and the third label curve of the current training record into a preset third model loss function; and based on a preset third model optimizer, performing a round of modulation on the model parameters of the basic power load forecasting model in a direction to minimize the third model loss function; Wherein, the third model loss function includes at least an L1 loss function, an L2 loss function and a SmoothL1 loss function; the third model optimizer includes at least an SGD optimizer and an Adam optimizer; Step 84, identifying whether the current training record is the last third data record of the third training set; if so, going to step 85; if not, taking the next third data record of the third training set as the new current training record and returning to step 83; Step 85, perform a round of traversal on all the third data records of the third evaluation set; and in this round of traversal, use the third data record currently traversed as the corresponding current evaluation record; and input the third training curve of the current evaluation record into the basic power load prediction model to perform energy storage charge and discharge curve prediction processing to obtain the corresponding sixth prediction curve; and form a corresponding third prediction-label pair from the sixth prediction curve and the third label curve of the current evaluation record; and at the end of this round of traversal, bring all the obtained third prediction-label pairs into the preset third model evaluation function to calculate and obtain the corresponding third evaluation value; Wherein, the third model evaluation function at least includes RMSE function; Step 86, identifying whether the third evaluation value meets the preset third evaluation value range; if not, returning to step 82 to continue training; if so, stopping training and confirming that the model training of the basic power load prediction model is completed.

9. The charging load control system of the substation according to claim 5, characterized in that: The performing model training on the charging load prediction model based on the fourth data set specifically includes: Step 91, based on a preset fourth segmentation ratio, the fourth data set is divided into two sub-data sets which are recorded as a corresponding fourth training set and a fourth evaluation set; Wherein, the fourth training set and the fourth evaluation set are both composed of a plurality of the fourth data records; the ratio of the total number of records of the fourth training set and the fourth evaluation set satisfies the fourth segmentation ratio; Step 92, taking the first fourth data record of the fourth training set as the corresponding current training record; Step 93, inputting the fourth training curve of the current training record into the charging load prediction model to perform charging load curve prediction processing to obtain the corresponding seventh prediction curve; and bringing the seventh prediction curve and the fourth label curve of the current training record into a preset fourth model loss function; and based on the preset fourth model optimizer, performing a round of modulation on the model parameters of the charging load prediction model in the direction of minimizing the fourth model loss function; Wherein, the fourth model loss function includes at least an L1 loss function, an L2 loss function and a SmoothL1 loss function; the fourth model optimizer includes at least an SGD optimizer and an Adam optimizer; Step 94, identifying whether the current training record is the last fourth data record of the fourth training set; if so, going to step 95; if not, taking the next fourth data record of the fourth training set as the new current training record and returning to step 93; Step 95, perform a round of traversal on all the fourth data records of the fourth evaluation set; and in this round of traversal, use the fourth data record currently traversed as the corresponding current evaluation record; and input the fourth training curve of the current evaluation record into the charging load prediction model to perform charging load curve prediction processing to obtain the corresponding eighth prediction curve; and form a corresponding fourth prediction-label pair by the eighth prediction curve and the fourth label curve of the current evaluation record; and at the end of this round of traversal, bring all the fourth prediction-label pairs obtained into the preset fourth model evaluation function to calculate and obtain the corresponding fourth evaluation value; Wherein, the fourth model evaluation function at least includes RMSE function; Step 96, identifying whether the fourth evaluation value meets the preset fourth evaluation value range; if not, returning to step 92 to continue training; if so, stopping training and confirming that the model training of the charging load prediction model is completed.

10. The charging load control system of the substation according to claim 1, characterized in that: The station area fusion terminal is specifically used for the charging total load prediction curve according to The charging control curves of all the charging piles in the current distribution area in the next period are set by using the charging available power curve △P to obtain the corresponding M charging pile control curves hour: The rated charging power of each charging pile in the station area is recorded as the corresponding rated charging power And for the M rated charging power Perform the sum calculation to get the corresponding rated power sum The total charging load prediction curve Divide the time period into N time periods on average to obtain a corresponding plurality of segmented curves, where N is a preset positive integer; perform mean square error calculation on each of the segmented curves to obtain a corresponding segmented mean square error; perform calculation on the absolute difference between each two segmented mean square errors to obtain a corresponding pairwise absolute difference; perform statistics on the number of pairwise absolute differences exceeding a preset absolute difference threshold to obtain a corresponding first number; and identify whether the first number is greater than 1; if not, set the corresponding control mode to a static mode; if so, set the corresponding control mode to a dynamic mode; the control mode includes a static mode and a dynamic mode; and identifying the obtained regulation mode; If the control mode is the static mode, the minimum power on the charging available power curve ΔP is used as the corresponding upper limit of the total available power p max ; Based on the total available power upper limit p max , the total rated power and all the rated charging powers stated Set the corresponding M electric pile control curves If the control mode is the dynamic mode, based on the charging available power curve ΔP and all the rated charging powers Set the corresponding M electric pile control curves 11. The charging load control system of the substation according to claim 10, characterized in that: The station area fusion terminal is specifically used for the max , the total rated power and all the rated charging powers stated Set the corresponding M electric pile control curves When each of the electric pile control curves The setting method is: Wherein, t is the control curve of the electric pile The time point information, is the electric pile control curve The power information corresponding to each of the time point information t.

12. The charging load control system of the substation according to claim 10, characterized in that: The station area fusion terminal is specifically used for the charging available power curve ΔP and all the rated charging power Set the corresponding M electric pile control curves When each of the electric pile control curves The setting method is: Wherein, t is the control curve of the electric pile The time point information, is the electric pile control curve The power information corresponding to each of the time point information t.

13. The charging load control system of the substation according to claim 1, characterized in that: The charging pile in the area is specifically used for controlling the process power of the charging process of the current charging vehicle under the guidance of the current charging control curve: Step 131, taking the access time of the current charging vehicle as the corresponding current start time; Step 132, extracting the control curve after the current start time on the current charging control curve as the corresponding current charging curve; and continuously charging the currently charged vehicle based on the current charging curve; Step 133, during the continuous charging process based on the current charging curve, real-time monitoring is performed on whether the charge state of the currently charged vehicle has reached the target specified by the owner of the currently charged vehicle, and real-time monitoring is performed on whether the current time has reached the end time of the current charging curve; and when the charge state of the currently charged vehicle reaches the target specified by the owner, the process goes to step 134; and when the current time has reached the end time of the current charging curve, the time period index corresponding to the current time is confirmed again based on the mapping relationship between the time period index and time to obtain a new current time period index, and the charging control curve Pc matching the time period index j in the multiple control curves stored locally in the charging pile with the new current time period index is matched. x,i,j As the new current charging control curve, and return to step 132 to enter the next continuous charging process; Step 134, stopping charging of the currently charging vehicle.

Images