CN117077381A

CN117077381A - Method for improving battery utilization efficiency by battery scheduling algorithm based on big data

Info

Publication number: CN117077381A
Application number: CN202310935241.4A
Authority: CN
Inventors: 蔡钺; 程禹斯; 章群华
Original assignee: Shanghai Zhizu Wulian Technology Co ltd
Current assignee: Shanghai Zhizu Wulian Technology Co ltd
Priority date: 2023-07-28
Filing date: 2023-07-28
Publication date: 2023-11-17

Abstract

The invention discloses a method for improving the utilization efficiency of a battery by a battery scheduling algorithm based on big data, which sequentially comprises the steps of data reporting and acquisition, data collection and storage, data calculation, data feedback, work order dispatch and battery scheduling execution; the method comprises the steps of establishing a gray scale model, further predicting future renting, exiting and replacing the battery cabinet by using the trained gray scale model, guiding battery allocation and scheduling according to the optimized gray scale model, generating scheduling data corresponding to a work order, sending a first-line scheduling person at the nearest position of the battery cabinet needing battery scheduling by the work order, leading the first-line scheduling person to the position of the battery cabinet appointed by the work order, displaying the allocation condition of the battery in the battery cabinet according to the information of the work order, and executing battery scheduling operation by combining a scheduling path. According to the invention, the power change order and the number of the batteries in the power change cabinet are monitored and calculated in real time, and the scheduling is optimized, so that the power change requirement of a rider is met, and the utilization efficiency of the batteries is improved.

Description

Method for improving battery utilization efficiency by battery scheduling algorithm based on big data

Technical Field

The invention belongs to the technical field of battery power-changing scheduling of two-wheelers, and particularly relates to a method for improving battery utilization efficiency by a battery scheduling algorithm based on big data.

Background

Today, takeouts have become a normal lifestyle of people, and two-wheel vehicle power conversion has become a popular energy supplement among the rider population in the takeouts. However, in the two-wheel vehicle electricity changing enterprises, the electricity changing experience of a rider is affected due to insufficient quantity of batteries in the electric cabinet, or the utilization efficiency of the batteries is low due to too many batteries in the electric cabinet, so that the electricity changing enterprises cannot be profitable, and the two-wheel vehicle electricity changing enterprises are serious and common problems generally faced by the two-wheel vehicle electricity changing enterprises at present.

Disclosure of Invention

The invention aims to: in order to overcome the defects in the prior art, the invention provides a method for improving the utilization efficiency of a battery by using a battery scheduling algorithm based on big data.

The technical scheme is as follows: in order to achieve the above purpose, the method for improving the battery utilization efficiency by using the big data-based battery scheduling algorithm of the invention comprises the following steps:

step S1: data reporting and acquisition

The method comprises the steps of detecting and calculating the number of batteries in a battery replacement cabinet and a battery replacement order in real time through equipment with the function of the Internet of things, and transmitting data to a big data platform;

step S2: data collection and storage

The data collection and storage are completed by adopting a data real-time access or offline synchronization mode;

step S3: data computation

After cleaning the data, establishing a gray scale model, predicting the number of people for starting renting, returning groups and changing the power cabinet in the future by using the trained gray scale model, guiding the battery to distribute and schedule according to the optimized gray scale model, and generating scheduling data corresponding to the work order;

step S4: data backhaul

The dispatching data is transmitted back to the work order system, the work order is directly established through a work order reading big data platform interface, and an instruction is automatically sent to a first-line battery dispatching personnel;

step S5: work order dispatch and battery dispatch execution

And after the position information of the battery changing cabinet, the position information of the first-line scheduling personnel and the distribution condition of the batteries in the battery changing cabinet are obtained, and an optimal scheduling path is determined by utilizing a path planning algorithm, the work order is automatically distributed to the first-line scheduling personnel at the position closest to the battery changing cabinet needing battery scheduling, the first-line scheduling personnel goes to the position of the battery changing cabinet designated by the work order, and the battery scheduling operation is executed according to the distribution condition of the batteries in the battery changing cabinet displayed by the work order information and in combination with the scheduling path.

Further, in step S1, the specific steps of reporting and collecting data are as follows:

a) The equipment collects and stores the data locally;

b) The device is connected to the cloud platform through a 2G/3G/4G/NB-IoT network;

c) The equipment establishes communication connection with a cloud platform;

d) The device uploads the data to the cloud platform and uses the MQTT communication protocol for communication;

e) The cloud platform receives, analyzes and stores the data uploaded by the equipment;

f) The platform distributes the data to the large data platform center via kafka.

Further, in step S3, the data calculation includes the steps of data cleaning, gray model building, model fusion and optimization, model evaluation and optimization, battery allocation policy, battery scheduling optimization, scheduling execution and feedback, performance evaluation and optimization in sequence, and specifically includes the following steps:

1) Data cleansing

Cleaning and preprocessing the collected original data, removing noise, processing abnormal values, and carrying out data format conversion and unification;

2) Gray model establishment

2.1 Determining the type of gray model: utilizing a GM (1, 1) model for capturing overall trends of the data;

2.2 Gray modeling is carried out on the historical data, and parameter estimation and simulation fitting are carried out;

2.3 Model training and prediction:

the original data sequence is rented by the site uploads: x≡0) = [ x≡0 (1), x≡0 (2),. The term x+0 (n) ], wherein: n represents historical days, X (0) represents the sequence of the last person renting number in the past n time periods, and X (0) (n) represents the specific last person renting number in the nth time period;

accumulating and generating a sequence: x≡1) = [ x≡1 (1), x≡1 (2), x+1 (n) ], wherein: x (1) (k) = Σx (0) (i), i from 1 to k

The accumulation and subtraction generating sequence: x≡2) = [ x≡2 (1), x≡2 (2) (2),. The x+2 (n) ], wherein: x (2) (k) = (x (1) (k) +x (1) (k+1))/2, k being from 1 to n-1;

gray amount sequence: z≡1) = [ z≡1 (2), z≡1 (3),. Z++1 (n) ], wherein: z (1) (k) =

-x (2) (k+1), k ranging from 1 to n-1;

gray differential equation: z (1) (k) +a x (1) (k) =b, wherein: a and b are parameters of a gray differential equation, and are estimated by a least square method;

the prediction model obtained by solving the gray differential equation is as follows: x (0) (k+1) = (x (1) (1) -b/a) exp (-a x k) +b/a), wherein: k is the number of prediction steps, x (0) (k+1) is the predicted value;

training the gray scale model by using the historical data in n time periods as a training set to obtain parameters of the model, and predicting the number of users who rent and change the cabinet in the future by using the trained gray scale model;

3) Model fusion and optimization

3.1 Collecting time sequence data of the battery changing cabinet and calculating an average value and a standard deviation;

3.2 Calculating the ranges of three standard deviations according to the average value and the standard deviation, wherein the upper limit is the standard deviation of the average value plus three times, and the lower limit is the standard deviation of the average value minus three times;

3.3 Checking whether the time series data falls within the range, and if the data exceeds the range, considering the abnormal condition;

fusing the prediction result of the gray level model with the prediction result of the normal distribution 3-sigama model;

GM (1, 1) predicted value: x (0) (t+1) = (x (1) (1) -b/a) exp (-a x t) +b/a)

The model fusion method is used, the prediction capability and the weight of the two models are comprehensively considered, and the fused prediction result is used as the final prediction result of the number of users who rent, rent and change the electric cabinet;

4) Model evaluation and optimization

Comparing the prediction result of the fusion model with the true value, calculating an evaluation index, a root mean square error and an average absolute error, and evaluating the prediction precision and accuracy of the model;

true value sequence: y= [ Y (1), Y (2), Y (T) ];

predicted value sequence of fusion model: y_pred= [ y_pred (1), y_pred (2), y_pred (T) ].

Root mean square error: rmse=sqrt (mean ((Y-y_pred)/(2));

average absolute error: mae=mean (abs (Y-y_pred));

wherein: mean () represents averaging, sqrt () represents square root operation, abs () represents absolute value operation;

and (5) optimizing and adjusting the model according to the evaluation result.

5) Battery allocation strategy

According to the real-time state and demand analysis of the battery in the battery changing cabinet, a battery allocation strategy is formulated, and the method specifically comprises the following steps:

5.1 Marking the fully charged battery as a usable battery, and marking the battery to be charged as a state to be charged;

5.2 Selecting a part of the batteries to be charged for charging according to the demand and the number of the available batteries so as to meet the demand of order taking;

5.3 When the number of available batteries is insufficient, determining the number of recoverable batteries according to the predicted lease-back order, and marking the recoverable batteries as available batteries;

6) Battery scheduling optimization

The battery is scheduled through an optimization algorithm to improve the utilization efficiency of the battery, and the method specifically comprises the following steps: the simulated annealing algorithm is used for carrying out optimized scheduling on the batteries, balancing the distribution of the batteries, balancing the charge and discharge times of the batteries and prolonging the service life of the batteries, and the method comprises the following steps of:

6.1 Battery allocation equalization target function):

assume that the battery state S is the remaining capacity: f_balance=abs (mean (S) -max (S));

charge-discharge number of times balance degree objective function:

assume that the battery state S is the remaining capacity: f_cycles=std (S);

comprehensively optimizing an objective function:

f_total=w1×f_balance+w2×f_cycles, wherein: w1 and w2 are weight coefficients for balancing the importance of the two optimization objectives;

6.2 Defining a state space and a scheme of a simulated annealing algorithm:

state definition: the state represents a battery scheduling scheme, represented by a binary code of length N:

I＝[i(1),i(2),...,i(N)],i(j)∈{0,1}

initial temperature: setting an initial temperature for controlling the probability of acceptance in the search process:

P(I)＝F(I)/Σ(F(I))

objective function: defining an objective function to evaluate the quality of the current state;

neighborhood search: generating an adjacent state through a mutation operation according to the current state, wherein the mutation operation is to perform random bit overturning or random bit replacement operation on the current state, and generating a neighborhood state I';

acceptance criteria: according to the Metropolis criterion, a poor state is accepted, avoiding trapping in a locally optimal solution:

p=exp ((F (I) -F (I'))/T, where T is the current temperature;

cooling strategy: gradually converging the searching process by controlling the temperature in a mode of reducing;

and (3) linearly cooling: t=α×t, where α is a cooling coefficient;

repeating the steps until the maximum iteration times are reached;

7) Scheduling execution and feedback

According to the optimized dispatching result, executing the battery allocation and dispatching strategy, specifically: the dispatching result is fed back to the battery changing cabinet management system, the command is automatically sent to a first-line battery dispatcher through the work order system for actual operation, and the distribution and dispatching conditions of the batteries are monitored and recorded for subsequent analysis and improvement;

8) Performance evaluation and optimization

And evaluating the performance of the optimal scheduling algorithm, comparing the difference between the actual scheduling result and the expected effect, optimizing and improving the algorithm according to the evaluation result, adjusting the scheduling strategy, and improving the battery utilization efficiency and the scheduling accuracy.

Further, in step S4, the specific steps of data backhaul are as follows:

the first step: model data transmission and work order state synchronization

Logic 1: when the model is evaluated and optimized, data are generated, are saved to mysql after being cleaned again, and are sent to a message center;

logic 2: a monitoring message center designates a queue, and real-time synchronization is carried out aiming at the status of the produced work order data;

and a second step of: creation and refreshing of worksheets

Logic 1: monitoring a message center designated queue, and creating and refreshing according to site differentiated work orders;

and a third step of: work order data feedback

Logic 1: by means of timing synchronization, the work order data is backed up to a big data calculation engine according to date dimension, and basic data is provided for model training

Further, in step S5, the specific steps of work order dispatch and battery scheduling are performed as follows:

1) Data preparation

Acquiring position information of a battery changing cabinet, position information of a first-line dispatcher and distribution conditions of batteries in the battery changing cabinet;

2) Path planning

Determining an optimal scheduling path according to the position of the battery changing cabinet and the position of a first-line scheduling person by utilizing a path planning algorithm formed by combining a shortest path algorithm with a genetic algorithm;

position of the battery changing cabinet: c= [ (x 1, y 1), (x 2, y 2), (x n, yn) ], representing the coordinate positions of n converter cabinets;

dispatcher position: p= [ (x 1, y 1), (x 2, y 2), (x m, ym) ], representing the coordinate positions of m first-line schedulers;

genetic algorithm means:

individual representation: the chromosomes are represented as an array, representing a dispatch path;

individuals: i= [ I (1), I (2), I (n) ], I (j) e {1,2,., n }

Objective function: the quality degree of a scheduling path, namely the path length, is measured;

objective function: f (I) =d (I (1), I (2)) +d (I (2), I (3)) + d (I (n-1), I (n)) +

Selecting:

selection probability: p (I) =f (I)/Σ (F (I))

Crossing:

sequentially crossing: selecting a random subsequence, and sequentially crossing the two parent individuals I1 and I2 to generate two child individuals I1 'and I2';

I1'＝[i(1),...,i(k),i'(k+1),...,i'(n)]

I2'＝[i'(1),...,i'(k),i(k+1),...,i(n)]

variation:

crossover variation: performing a swap operation on two locations in the child generation;

3) Automatic dispatch worksheet

3.1 Establishing an interface with a work order system, and automatically distributing a path planning result to first-line scheduling personnel through data transmission;

3.2 The work order system generates a work order containing detailed information of the scheduling task according to the path planning result and automatically sends the work order to corresponding first-line scheduling personnel;

4) Real time monitoring and feedback

4.1 A front-line dispatcher can receive and check the dispatched work order through the mobile terminal equipment;

4.2 The work order system monitors the position information of a first-line dispatcher in real time so as to track the execution progress of the work order and the position of the dispatcher;

4.3 After completing the task scheduling, the on-line scheduling personnel feeds back the task completion condition on the mobile terminal equipment.

The beneficial effects are that: the beneficial effects of the invention are as follows:

(1) The electricity exchanging experience of the user is ensured, and the electricity taking-out quantity is improved;

(2) The asset utilization efficiency of the battery is improved, cash is saved, and profit is improved;

(3) The dispatching efficiency and the dispatching quantity of the dispatching personnel in unit time are improved.

Drawings

FIG. 1 is a block diagram of the method of the present invention;

FIG. 2 is a block diagram of the data reporting and collecting structure;

FIG. 3 is a block diagram of a real-time access scheme in data collection and storage;

FIG. 4 is a block diagram of an offline synchronization scheme in data collection and storage;

fig. 5 is a block diagram of a data backhaul.

Detailed Description

The invention will be further described with reference to the accompanying drawings.

As shown in fig. 1, a method for improving battery utilization efficiency by using a battery scheduling algorithm based on big data comprises the following steps:

step S1: data reporting and acquisition

And detecting and calculating the number of batteries in the battery exchange cabinet and the battery exchange order in real time through equipment with the function of the Internet of things, and transmitting data to a big data platform.

As shown in fig. 2, in step S1, the specific steps of data reporting and collecting are as follows:

a) The equipment collects and stores the data locally;

c) The equipment establishes communication connection with a cloud platform;

Step S2: data collection and storage

And data collection and storage are completed by adopting a data real-time access or offline synchronization mode.

As shown in fig. 3, the real-time access mode

Data format:

attributes of	Paraphrasing meaning	Remarks
			imei	IEMI of device
protocol_type	Reporting protocol type
			message	Device reporting content	Unprocessed messages
topic	Device heartbeat source

1. The data center confirms the reporting protocol and the message channel through developing and configuring the Flink task;

2. the data center receives the push data of the corresponding kafka;

3. the data center stores the original data;

4. and the data center analyzes the received message according to a preset protocol to finish the storage process.

As shown in FIG. 4, the off-line synchronization mode

1. Configuring a data source and a target source: configuring synchronous data sources and target sources in a configuration center, wherein the synchronous data sources and target sources comprise data source types, connection information, table names and the like;

2. defining a synchronization task: defining a data source and a target source to be synchronized, a data column, a synchronization start-stop position and the like according to service requirements;

3. generating a data synchronization script: generating a corresponding data synchronization script according to the task definition;

4. performing data synchronization: and executing data synchronization operation according to the generated script, and synchronizing the appointed source data into the target data source.

Step S3: data computation

And after the data are cleaned, a gray scale model is established, the trained gray scale model is utilized to predict the number of people who rent, withdraw a group and use the power conversion cabinet in the future, and the battery is guided to be distributed and scheduled according to the optimized gray scale model to generate scheduling data corresponding to the work order.

In step S3, the data calculation includes the steps of data cleaning, gray model building, model fusion and optimization, model evaluation and optimization, battery allocation policy, battery scheduling optimization, scheduling execution and feedback, performance evaluation and optimization in sequence, and specifically includes the following steps:

1) Data cleansing

And cleaning and preprocessing the collected original data, removing noise, processing abnormal values, and performing data format conversion and unification.

Processing the missing values:

detecting a missing value: checking the data to find out the field with the missing value;

missing value filling: according to the data reported by the battery-changing cabinet, the missing value is filled with numerical data by using the mean value, the median value and the mode value under the condition that the historical behavior data quantity of the user is large, and the missing value is filled with the data by using a multiple interpolation method Micefrest under the condition that the historical behavior data quantity of the user is small.

Processing outliers:

abnormal value detection: detecting abnormal values in the data by a statistical analysis method;

outlier processing: for the data determined to be abnormal values, if a large number of abnormal values exist, deletion is selected, and other cases are processed by using a replacement/interpolation method.

Data format conversion and unification:

data type conversion: converting the data field into proper data type according to the meaning and analysis requirement of the data, such as converting the date field into date format, and converting the character string field into numeric or category type data;

and (3) unified treatment: and carrying out unit unification processing on the data fields containing the units to ensure the consistency of the data.

And (3) data verification:

and (3) logic verification: checking logic rules of the data, such as numerical ranges, consistency of associated fields and the like;

integrity verification: ensuring data integrity, such as checking of mandatory fields, primary key uniqueness, etc.

Numerical value type feature extraction:

standardization: for numerical features, a normalization process may be performed to convert them to a standard normal distribution with zero mean and unit variance;

and (5) barrel separation: the continuous type feature is divided into a plurality of discrete intervals, which are converted into the category type feature.

Category type feature extraction:

single heat coding: for class type features with limited values, converting by adopting One-Hot Encoding (One-Hot Encoding), wherein the One-Hot Encoding converts each class into a binary feature vector, and only One position is 1, and the rest positions are 0;

ordered coding: for class type features with natural ordering relations, orderly Encoding (Ordinal Encoding) is used to convert the class type features into continuous numerical type features, and the order information of the class type features is reserved.

And (3) extracting time characteristics:

timestamp processing: for the characteristics containing time information, performing time stamp processing, and extracting relevant time characteristics such as year, quarter, month, day of week, hour and the like;

calculating a time difference: the time difference between the time features, the duration of the order, etc. may be calculated.

Extracting geographic information characteristics:

and (3) calculating the distance: for the position information of the power exchange cabinets, calculating the distance between the two power exchange cabinets according to the longitude and latitude, and taking the distance as a characteristic;

geocoding: and converting the geographical position information into category type features such as region codes, administrative division and the like.

Extracting features of domain expertise:

relevant characteristics are extracted according to professional knowledge in the problem field, such as the characteristics related to the health state of the battery according to factors such as the charging characteristics, the temperature and the like of the battery.

2) Gray model establishment

2.3 Model training and prediction:

-x (2) (k+1), k ranging from 1 to n-1;

3) Model fusion and optimization

GM (1, 1) predicted value: x (0) (t+1) = (x (1) (1) -b/a) exp (-a x t) +b/a)

Using a model fusion (Stacking, ensembling) method, comprehensively considering the prediction capability and the weight of the two models, and taking the fused prediction result as the final prediction result of the number of users who rent, rent and change the electric cabinet;

4) Model evaluation and optimization

Comparing the prediction result of the fusion model with a true value, calculating an evaluation index, a Root Mean Square Error (RMSE) and a Mean Absolute Error (MAE), and evaluating the prediction precision and accuracy of the model;

true value sequence: y= [ Y (1), Y (2), Y (T) ];

Root mean square error: rmse=sqrt (mean ((Y-y_pred)/(2));

average absolute error: mae=mean (abs (Y-y_pred));

and (3) optimizing and adjusting the model according to the evaluation result, such as adjusting parameters, improving the model structure and the like.

5) Battery allocation strategy

6) Battery scheduling optimization

6.1 Battery allocation equalization target function):

charge-discharge number of times balance degree objective function:

assume that the battery state S is the remaining capacity: f_cycles=std (S);

comprehensively optimizing an objective function:

6.2 Defining a state space and a scheme of a simulated annealing algorithm:

I＝[i(1),i(2),...,i(N)],i(j)∈{0,1}

P(I)＝F(I)/Σ(F(I))

objective function: defining an objective function to evaluate the quality of the current state, for example, considering the distribution balance degree and the charge and discharge frequency balance degree of the battery;

p=exp ((F (I) -F (I'))/T, where T is the current temperature;

and (3) linearly cooling: t=α×t, where α is a cooling coefficient;

repeating the steps until the maximum iteration times are reached;

in the dispatching process, factors such as capacity, charging speed, discharging speed and the like of the battery are considered, so that the demand of taking a renting order is met to the greatest extent, and the idle time and the waiting time of the battery are reduced.

7) Scheduling execution and feedback

8) Performance evaluation and optimization

Step S4: data backhaul

And the dispatching data is transmitted back to the work order system, the work order is directly created by reading the large data platform interface through the work order, and the instruction is automatically sent to a line battery dispatcher.

As shown in fig. 5, in step S4, the specific steps of data backhaul are as follows:

the first step: model data transmission and work order state synchronization

and a second step of: creation and refreshing of worksheets

and a third step of: work order data feedback

Step S5: work order dispatch and battery dispatch execution

In step S5, the specific steps of work order dispatch and battery scheduling are as follows:

1) Data preparation

2) Path planning

genetic algorithm means:

individuals: i= [ I (1), I (2), I (n) ], I (j) e {1,2,., n }

Selecting:

selection probability: p (I) =f (I)/Σ (F (I))

Crossing:

I1'＝[i(1),...,i(k),i'(k+1),...,i'(n)]

I2'＝[i'(1),...,i'(k),i(k+1),...,i(n)]

variation:

crossover variation: performing a swap operation on two locations in a sub-generation individual, wherein the path length is represented by calculating a euclidean distance or other distance metric between two coordinate points, such as d (i, j) =sqrt ((xi-xj)/(2+ (yi-yj)/(2));

the working capacity and efficiency of the on-line scheduling personnel and the distribution condition of the batteries in the battery changing cabinet are considered, so that the path planning can complete the scheduling task as quickly as possible.

3) Automatic dispatch worksheet

3.2 The work order system generates a work order containing detailed information of the scheduled tasks according to the path planning result and automatically sends the work order to corresponding first-line scheduling personnel, wherein the detailed information of the scheduled tasks is as the position of a power change cabinet to be scheduled, the number of batteries to be scheduled, the destination and the like.

4) Real time monitoring and feedback

4.3 After completing the task scheduling, the on-line scheduling personnel feeds back the task completion condition on the mobile terminal equipment, including the information such as task execution time, battery state and the like.

The invention has the advantages that:

In conclusion, the invention monitors and calculates the power change order and the battery quantity in the power change cabinet in real time, and optimizes the scheduling to meet the power change requirement of a rider, and improves the utilization efficiency of the battery.

In addition, user experience can be improved, waiting time is shortened, energy waste is reduced, and the power change of the two-wheel vehicle is more intelligent and efficient.

The foregoing is only a preferred embodiment of the invention, it being noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the present invention, and such modifications and adaptations are intended to be comprehended within the scope of the invention.

Claims

1. A method for improving battery utilization efficiency by a battery scheduling algorithm based on big data is characterized by comprising the following steps: the method comprises the following steps:

step S1: data reporting and acquisition

step S2: data collection and storage

step S3: data computation

step S4: data backhaul

step S5: work order dispatch and battery dispatch execution

2. The method for improving the battery utilization efficiency by using the big data-based battery scheduling algorithm according to claim 1, wherein the method comprises the following steps of: in step S1, the specific steps of data reporting and collecting are as follows:

a) The equipment collects and stores the data locally;

c) The equipment establishes communication connection with a cloud platform;

3. The method for improving the battery utilization efficiency by using the big data-based battery scheduling algorithm according to claim 1, wherein the method comprises the following steps of: in step S3, the data calculation includes the steps of data cleaning, gray model building, model fusion and optimization, model evaluation and optimization, battery allocation policy, battery scheduling optimization, scheduling execution and feedback, performance evaluation and optimization in sequence, and specifically includes the following steps:

1) Data cleansing

2) Gray model establishment

2.3 Model training and prediction:

-x (2) (k+1), k ranging from 1 to n-1;

3) Model fusion and optimization

GM (1, 1) predicted value: x (0) (t+1) = (x (1) (1) -b/a) exp (-a x t) +b/a)

4) Model evaluation and optimization

true value sequence: y= [ Y (1), Y (2), Y (T) ];

Root mean square error: rmse=sqrt (mean ((Y-y_pred)/(2));

average absolute error: mae=mean (abs (Y-y_pred));

and (5) optimizing and adjusting the model according to the evaluation result.

5) Battery allocation strategy

6) Battery scheduling optimization

6.1 Battery allocation equalization target function):

charge-discharge number of times balance degree objective function:

assume that the battery state S is the remaining capacity: f_cycles=std (S);

comprehensively optimizing an objective function:

6.2 Defining a state space and a scheme of a simulated annealing algorithm:

I＝[i(1),i(2),...,i(N)],i(j)∈{0,1}

P(I)＝F(I)/Σ(F(I))

p=exp ((F (I) -F (I'))/T, where T is the current temperature;

and (3) linearly cooling: t=α×t, where α is a cooling coefficient;

repeating the steps until the maximum iteration times are reached;

7) Scheduling execution and feedback

8) Performance evaluation and optimization

4. The method for improving the battery utilization efficiency by using the big data-based battery scheduling algorithm according to claim 1, wherein the method comprises the following steps of: in step S4, the specific steps of data backhaul are as follows:

the first step: model data transmission and work order state synchronization

and a second step of: creation and refreshing of worksheets

and a third step of: work order data feedback

Logic 1: and backing up the work order data to a big data calculation engine according to the date dimension in a timing synchronization mode, and providing basic data for model training.

5. The method for improving the battery utilization efficiency by using the big data-based battery scheduling algorithm according to claim 1, wherein the method comprises the following steps of: in step S5, the specific steps of work order dispatch and battery scheduling are as follows:

1) Data preparation

2) Path planning

genetic algorithm means:

individuals: i= [ I (1), I (2), I (n) ], I (j) e {1,2,., n }

Selecting:

selection probability: p (I) =f (I)/Σ (F (I))

Crossing:

I1'＝[i(1),...,i(k),i'(k+1),...,i'(n)]

I2'＝[i'(1),...,i'(k),i(k+1),...,i(n)]

variation:

3) Automatic dispatch worksheet

4) Real time monitoring and feedback