New energy power intelligent allocation method

Abstract

The application provides a new energy power intelligent allocation method, comprising the following steps: extracting the power load change trend of the teaching area and the industry-university-research area according to the historical power consumption data of the campus and the industry-university-research area, analyzing the power consumption cycle difference between the holiday and the school season, and determining the power consumption complementary time period; based on the power consumption plan data of the campus and the industry-university-research area, the deployment of the photovoltaic array dynamic strategy is carried out, the target deployment position of the photovoltaic array in different time periods is calculated by combining the moving ability and the power generation efficiency of the photovoltaic array, and the dynamic deployment scheme of the photovoltaic array is generated; according to the actual execution data of the photovoltaic array dynamic deployment scheme, the comparison result of the photovoltaic power generation capacity and the power consumption demand is obtained, the effect of the power coordination and mutual aid is analyzed, the photovoltaic array deployment strategy and the peak-shaving production mechanism are optimized, and the school-enterprise linkage power supply scheme is generated.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, in particular to a new energy power intelligent deployment method. BACKGROUND

[0002] The power demand of the campus teaching area and the production-study-research area is obviously different in different periods, how to use the movable photovoltaic array to realize the dynamic balance and mutual aid of the power of both sides is a complex technical problem. The power demand of the campus and the production-study-research area in different periods has obvious complementary characteristics, the teaching area has high power load in the semester, and the load obviously decreases during the holiday, and the power load of the production-study-research area is relatively stable.

[0003] How to adjust the layout scheme of the photovoltaic array in real time according to the power curves of both sides, realize the array transfer and reconstruction in the limited roof, playground, parking lot and other sites, and ensure the stable docking of the power transmission line, there are many technical bottlenecks. And when the school power load rises in the school season, the production scheduling scheme of the production-study-research area needs to be studied, the power support capacity of the school under the peak-shaving production mode is analyzed, and the dynamic adjustment of the photovoltaic array layout is considered again to match the load change trend, which puts high requirements on the mechanical mobility, deployment adaptability and grid stability of the array. Finally, the mobile deployment of the photovoltaic array will affect the power generation efficiency, how to need to deeply analyze the relationship between the array layout and the power generation efficiency, and study the dynamic deployment strategy of the array based on the power demand of both sides, meet the power demand while taking into account the power generation benefit, which is a technical problem to be solved. SUMMARY

[0004] The present application provides a new energy power intelligent deployment method, which can improve the energy utilization efficiency.

[0005] In order to achieve the above-mentioned purpose, the following technical scheme is provided.

[0006] The new energy power intelligent deployment method extracts the power load change trend of the teaching area and the production-study-research area according to the historical power data of the campus and the production-study-research area, analyzes the power cycle difference between the holiday and the school season, determines the power complementary time period, and calculates the power generation data of the photovoltaic array after moving according to the target position of the photovoltaic array moving.

[0007] The power generation data of the photovoltaic array after moving is calculated according to the target position of the photovoltaic array moving, combined with the power demand data of the production-study-research area, whether the photovoltaic power generation meets the power demand of the production-study-research area is judged, if not, the peak-shaving production mechanism is started to generate the deployment position of the photovoltaic array in the power load reduction stage of the teaching area.

[0008] The power generation data of the photovoltaic array after moving is calculated according to the target position of the photovoltaic array moving, combined with the power demand data of the production-study-research area, whether the photovoltaic power generation meets the power demand of the production-study-research area is judged, if not, the peak-shaving production mechanism is started to generate the deployment position of the photovoltaic array in the power load reduction stage of the teaching area.

[0009] Obtain the electricity plan data of the production-study-research zone, analyze the electricity load change trend of the production-study-research zone and determine the time period of peak load shifting, according to the characteristics of the electricity load of the teaching area rising during the school period, adjust the electricity plan of the production-study-research zone to support the electricity demand of the teaching area;

[0010] Based on the electricity plan data of the campus and the production-study-research zone, the dynamic strategy of the photovoltaic array is deployed, the target deployment position of the photovoltaic array in different time periods is calculated combined with the moving ability and power generation efficiency of the photovoltaic array, and the dynamic deployment scheme of the photovoltaic array is generated;

[0011] By comparing the dynamic deployment strategy scheme of the photovoltaic array with the electricity plan data, it is judged whether the photovoltaic power generation capacity meets the electricity demand of the campus and the production-study-research zone, if it meets, the deployment scheme is executed, if it does not meet, the deployment position of the photovoltaic array is adjusted or the peak load shifting mechanism is started;

[0012] According to the actual execution data of the dynamic deployment scheme of the photovoltaic array, the comparison result of the photovoltaic power generation capacity and the electricity demand is obtained, the effect of the power coordination and mutual aid is analyzed, the deployment strategy of the photovoltaic array and the peak load shifting mechanism are optimized, and the school-enterprise joint power supply scheme is generated.

[0013] The technical scheme provided by the embodiment of the present application can include the following beneficial effects:

[0014] The present application discloses a new energy power intelligent deployment method. The method analyzes the electricity load change trend of the campus teaching area and the production-study-research zone, combines the power generation efficiency of the photovoltaic array in different areas, and formulates the dynamic deployment strategy of the photovoltaic array. According to the electricity cycle difference between the holiday and the school season, the present application determines the electricity complementary time period and calculates the power generation potential of the photovoltaic array in different positions. By establishing the power coordination and mutual aid model, the present application judges whether the photovoltaic power generation capacity meets the electricity demand of each area, and starts the peak load shifting mechanism if necessary. The present application also optimizes the deployment strategy and the peak load shifting mechanism according to the actual execution data, and finally generates the school-enterprise joint power supply scheme, realizes the dynamic matching of photovoltaic power generation and electricity demand, and improves the energy utilization efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The flowchart of the new energy power intelligent deployment method of the present application. DETAILED DESCRIPTION

[0016] The present application will be described in detail below in combination with specific embodiments.

[0017] As Figure 1 , the new energy power intelligent deployment method of the present application can specifically include:

[0018] In step S101, according to historical power consumption data of the campus and the industry-university-research area, the power consumption load variation trend of the teaching area and the industry-university-research area is extracted, the power consumption cycle difference between the holiday and the school season is analyzed, and the power consumption complementary time period is determined.

[0019] The power consumption load curve collected by the online monitoring sensor is acquired, the missing data points in the power consumption load curve are repaired by using a moving average interpolation method, and a power consumption load distribution data table containing a time stamp, a load value and a peak-valley identifier is obtained; according to the power consumption load distribution data table, a power consumption load variation rate threshold is set, data points exceeding the threshold are marked, and a cycle feature table containing a time period, a load variation amount and a duration is obtained; for the cycle feature table, the power consumption load mean value, the standard deviation and the coefficient of variation are calculated by using a hierarchical clustering method, and an interval boundary time point sequence is obtained; according to the interval boundary time point sequence, a sliding time window method is used for segmented processing of equipment operation feature data, a dynamic programming algorithm is used for time dimension redistribution of power consumption load of each area, and a complementary interval load regulation scheme is obtained.

[0020] Specifically, the electricity load curve of the production-study-research area and the teaching area collected by the online monitoring sensor is obtained from the historical electricity record database. The missing data points are repaired by using the moving average interpolation method. The electricity load characteristic values are obtained by calculating the peak-valley difference, duration, and change rate of the electricity load curve. A distribution data table containing the timestamp, load value, and peak-valley identifier is generated. For the electricity load curve of the production-study-research area and the teaching area, the electricity load change rate threshold is set according to the electricity load distribution data table. The data points exceeding the threshold are marked. By comparing the marked points with the historical holiday and school opening time, a period characteristic table containing the time period, load change amount, and duration is generated. For the period characteristic table, the electricity complementary time interval is divided by using the hierarchical clustering method according to the mean, standard deviation, and coefficient of variation of the electricity load of the production-study-research area and the teaching area in different time periods. The interval boundary time point sequence is generated. According to the interval boundary time point sequence, the running status of the electricity equipment in each interval is counted. The equipment startup and shutdown time, rated power, and actual load are recorded. The equipment running characteristic data set is generated. The equipment running characteristic data set is processed by using the sliding time window method. The load distribution in each time window is calculated. The least squares method is used to fit the load change trend curve. According to the load change trend curve, the dynamic programming algorithm is used to redistribute the electricity load of each area in the time dimension, and the complementary interval load regulation scheme is generated. In the electricity load data collection stage, the monitoring sensor usually records the electricity consumption data at intervals of 15 minutes. When data is missing, the moving average interpolation method is used to repair the missing points. The mean of the previous and next time points is calculated to fill the data, ensuring the continuity of the data. The load curve characteristic values include the peak-valley difference, duration, and change rate. The peak-valley difference reflects the electricity fluctuation amplitude, the duration reflects the stability of the load state, and the change rate represents the load change speed. For a certain production-study-research area, the electricity peak is from 9 am to 17 pm, with a peak load of 300 kilowatt-hours. The electricity valley is from 22 pm to 6 am the next day, with a valley load of 50 kilowatt-hours, forming a typical day-high night-low load characteristic. In the period characteristic identification process, the load change rate threshold is set to 30%. The data points exceeding the threshold are marked. It is found that one week before the winter and summer vacations, the teaching area electricity load decreases from 250 kilowatt-hours to 75 kilowatt-hours, with a decrease of 70%. While the production-study-research area decreases from 280 kilowatt-hours to 210 kilowatt-hours, with a decrease of 25%, which reflects the significant influence of the school period on the electricity load of the teaching area. By calculating the mean and standard deviation of the electricity load in different time periods, the hierarchical clustering method is used to divide a day into 6 time intervals. The two electricity peak intervals are from 9 am to 11 am and from 14 pm to 16 pm. The mean load of the teaching area in this period is 280 kilowatt-hours, and the standard deviation is 15 kilowatt-hours, which reflects the relatively stable electricity characteristics.The production-study-research area presents a higher load from 11:00 to 14:00 and from 16:00 to 19:00, with an average of 260 kWh, forming a peak-load shifting mode with the teaching area. In the analysis of the equipment operation characteristics, it is found that the air conditioning equipment in the teaching building accounts for 45% of the total load, the lighting equipment accounts for 25%, and other equipment accounts for 30%. Using a 30-minute sliding time window to segment the equipment operation data, it is calculated that the load of the air conditioning equipment decreases by 40% after the temperature reaches the set value, and the load of the lighting equipment gradually changes with the natural light intensity. By least squares fitting the load variation trend of each type of equipment, a load prediction model is established. Combined with the power equipment capacity constraint, the time dimension optimization configuration of the power load of each area is carried out, and under the premise of meeting the teaching area course time requirement, part of the adjustable load in the production-study-research area is moved to the power low valley period, realizing the efficient use of power resources. Through the dynamic programming algorithm, the optimal scheduling scheme is calculated, so that the adjusted peak-valley difference rate is reduced from 85% to 60%, and the overall load curve of the system is more stable.

[0021] In step S102, the campus photovoltaic array power generation efficiency data is obtained, combined with the campus space layout information and the complementary power consumption time period, the photovoltaic array deployment scheme is updated, and the updated power generation potential is calculated, the characteristics of the power consumption load increase of the teaching area are determined, and the target position of the photovoltaic array movement is determined.

[0022] The array power generation data and the light intensity data collected by the photovoltaic monitor are obtained, the light resource distribution data table is obtained by Fourier transform of the light intensity data, the light resource distribution data table includes timestamp, light intensity, and power generation efficiency; according to the light resource distribution data table and the three-dimensional laser scanning data, the site boundary coordinate point set is extracted by using a deep learning segmentation algorithm, and the site layout constraint table is calculated by vector projection; according to the site layout constraint table and the power consumption load curve record, the power consumption load curve record is segmented and calculated by using a sliding window to obtain a power consumption characteristic data table; according to the power consumption characteristic data table and the site layout constraint table, a multi-objective genetic algorithm is used to optimize the solar panel arrangement to obtain a photovoltaic array deployment scheme, the photovoltaic array deployment scheme includes equipment number, installation position, and orientation angle.

[0023] Specifically, the hourly power generation data and ten-minute interval light intensity data of the array on the roof of the teaching area are collected from the photovoltaic monitor. For the current orientation angle and position coordinates of the solar panel, the light data is decomposed into a 24-hour cycle through Fourier transform to obtain a light resource distribution data table containing timestamp, light intensity, and power generation efficiency. The characteristics of the electrical load include power consumption, peak-valley power difference, daily average power consumption fluctuation, and power generation potential improvement. According to the three-dimensional laser scanning data of the playground and parking lot, the deep learning segmentation algorithm is used to extract the three-dimensional coordinate point set of the site boundary, calculate the building sunshine duration and shadow coverage area, and calculate the best inclination angle of the solar panel in different seasons through vector projection to generate a site layout constraint table containing spatial coordinates, inclination parameters, and shadow coverage rate. According to the light resource distribution data table and the site layout constraint table, a candidate installation area list is generated through data association matching, and the annual power generation prediction value of each area is calculated. According to the electrical feature data table and the candidate installation area list, a multi-objective genetic algorithm is used to optimize the solar panel layout scheme, with the optimization objectives set as maximizing power generation efficiency and minimizing shadow interference, to generate a photovoltaic array deployment scheme containing equipment number, installation location, and orientation angle. According to the photovoltaic array deployment scheme, combined with the building load constraints and site boundary restrictions, a device movement route map from the current location to the target location is generated through the path planning algorithm to obtain a device migration instruction set containing path nodes, turning angles, and movement distances. The power generation efficiency monitoring of the photovoltaic array records power generation data at the hourly level, and the light intensity value is collected every 10 minutes. This multi-dimensional data collection method helps to capture the impact of weather changes on power generation efficiency. The data records of a certain teaching area show that the light intensity reaches 1000 watts per square meter from 11am to 14pm on a sunny summer day, and the photovoltaic panel has the highest power generation efficiency, with a power generation capacity of 160 watt-hours per square meter. Through Fourier transform, the light data is decomposed into a 24-hour cycle to identify the fluctuation pattern of light resources and obtain power generation potential evaluation values for different time periods. The site layout analysis uses three-dimensional laser scanning technology to obtain spatial point cloud data of the playground and parking lot, with 500 sampling points per square meter recording spatial three-dimensional coordinate information. After extracting the site boundary using the deep learning segmentation algorithm, a three-dimensional model of the site is established to calculate the projection trajectory of the building. During the winter season from 8am to 16pm, the teaching building blocks the east side of the playground, affecting an area of 800 square meters, while the parking lot area is less affected. Through vector projection calculation, it is found that the best inclination angle of the photovoltaic panel in the parking lot area is 25 degrees in summer and 40 degrees in winter.The teaching area electricity load curve is segmented and counted by a 4-hour sliding window. It is found that during the summer vacation, the daily electricity consumption decreases from 2400 kWh during the teaching period to 800 kWh, and the peak-valley difference decreases from 1600 kWh to 400 kWh. The matching degree analysis of photovoltaic power generation and electricity load shows that the self-consumption ratio of photovoltaic power generation during the vacation period can be increased by 40%. The candidate installation area is selected according to the illumination conditions and site constraints. The parking area is 2000 square meters, and considering the equipment installation spacing and maintenance channel requirements, the actual available area is 1400 square meters. The multi-objective genetic algorithm is used to optimize the solar panel layout, the population size is set to 100, the iteration number is 500, the crossover probability is 0.8, and the mutation probability is 0.1. The optimization results show that by moving the existing photovoltaic array from the roof of the teaching building to the parking area and using a double-row arrangement, the annual power generation can be increased by 25%. The equipment moving route planning considers the site traffic restrictions and divides the relocation process into several sub-tasks. The vertical transportation channel from the roof of the teaching building to the ground is 3 meters wide, the turning radius is not less than 4 meters, and the horizontal transportation distance is 150 meters. The equipment relocation scheme generated by the path planning algorithm includes 12 key nodes, each node records the equipment position coordinates and turning angle, ensuring the safety of equipment transportation.

[0024] In step S103, according to the target position of the photovoltaic array movement, the power generation data after the photovoltaic array movement is calculated, and combined with the electricity demand data of the production-research-institute area, it is judged whether the photovoltaic power generation meets the electricity demand of the production-research-institute area. If not, the peak-shaving production mechanism is started to generate the photovoltaic array deployment position in the teaching area electricity load reduction stage.

[0025] The illumination intensity is calculated by Gaussian process regression according to the spatial coordinates of the photovoltaic array, the kernel function of the Gaussian process regression is selected as the radial basis function, and the photovoltaic panel power generation prediction data table is obtained; the random forest algorithm is used to decompose the building electricity load, the feature dimension of the random forest algorithm includes timestamp, temperature, and passenger flow data, and the sub-item electricity data table is obtained; the load classification result table is obtained by calculating the proportion of each type of load in the sub-item electricity data table by hierarchical clustering method, and the hierarchical clustering method sets the clustering distance threshold as the total load percentage parameter; for the power generation prediction data table and the load classification result table, the time sequence matching method is used to calculate the power supply gap value, and if the power supply gap value is greater than the preset threshold, the peak-shaving scheduling instruction set is generated.

[0026] Specifically, according to the spatial coordinate data of the photovoltaic array at the target position, the future light intensity and power generation are calculated using Gaussian process regression, the kernel function is selected as radial basis function, and the bandwidth parameter is set to 24 hours. Combined with the shadow blocking rate and the photoelectric conversion efficiency decay curve, the photovoltaic panel power generation prediction data table in each time period is obtained by trapezoidal integration method. The power monitoring unit obtains the electricity load curve of the research and application area, and the random forest algorithm is used to decompose the building electricity load, the decision tree depth is set to 8, and the feature dimension includes timestamp, temperature and passenger flow, and the sub-item electricity data table including lighting load, power load and air conditioning load is generated. According to the sub-item electricity data table, the proportion of each type of load is calculated by hierarchical clustering method, and the clustering distance threshold is set to 30% of the total load to obtain the load classification result table. For the photovoltaic power generation prediction data table and the load classification result table, the hourly power supply gap value is calculated by time sequence matching method, and if the power supply gap value is greater than the preset threshold, the peak shaving scheduling instruction set including equipment type, operation period and load value is generated. The integer programming method is used to optimize the layout of the photovoltaic array in the teaching area, and the constraint conditions include site area limit, equipment spacing requirement and building load limit, and the objective function is to maximize the power generation efficiency. According to the peak shaving scheduling instruction set and the photovoltaic array layout scheme, the load transfer scheme is calculated by dynamic programming algorithm, and the load balance data table including start and end time, load adjustment amount and equipment operation state is generated. The photovoltaic power generation prediction adopts Gaussian process regression method, which captures the periodic variation law of light intensity through radial basis kernel function, and the kernel function bandwidth parameter is set to 24 hours, corresponding to one day of light period. In practical application, the power generation efficiency of photovoltaic panel will gradually decrease with the use time, the conversion efficiency of newly installed photovoltaic panel is 18%, and after 5 years of use, it decreases to 16.2%. The actual power generation is 360 watts when the area of a single photovoltaic panel is 2 square meters and the light intensity is 1000 watts per square meter at noon on a sunny day. The building electricity load decomposition adopts random forest method, and the load data is classified through 8-layer decision tree structure. The input features include timestamp, temperature and passenger flow data, among which the temperature data comes from the weather station with an interval of 1 hour, and the passenger flow data comes from the access control system, recording the number of people passing through in 5 minutes. The lighting load proportion of a certain teaching building is 25% from 9 am to 17 pm on weekdays, the air conditioning load proportion is 45%, and the power load proportion is 30%. Load clustering analysis adopts hierarchical clustering method to group the electricity using equipment according to load characteristics. The clustering distance threshold is set to 30% of the total load to obtain three categories of high load, medium load and low load. Among them, the air conditioner, elevator and experimental equipment belong to the high load category, with a total capacity of 500 kilowatts; the lighting and computer equipment belong to the medium load category, with a total capacity of 300 kilowatts; and other office equipment belongs to the low load category, with a total capacity of 200 kilowatts. The power supply gap is calculated based on the time sequence matching principle, and the time distribution of photovoltaic power generation and electricity demand is compared.When the photovoltaic power generation is lower than the electricity demand, if the difference exceeds the preset threshold of 100 kilowatts, the peak clipping scheduling mechanism is triggered. The peak clipping scheduling instruction sets different delay times according to the load level, and the high-load equipment is delayed for 30 minutes, and the medium-load equipment is delayed for 60 minutes. The photovoltaic array arrangement optimization adopts an integer programming method, the site area limit is 2000 square meters, the minimum spacing between photovoltaic panels is 1.5 meters, and the building roof load limit is 60 kilograms per square meter. Under the condition of meeting these constraints, the optimal installation position and orientation angle are found. The optimization result shows that the power generation efficiency is the highest when the southward 25-degree inclination and double-row arrangement mode are adopted. The load balancing adjustment adopts a dynamic programming algorithm, and 24 hours are divided into 48 time periods, each time period being 30 minutes. By calculating the load transfer cost of each time period, an optimal scheduling scheme is generated. In the electricity peak period from 14 to 16 hours, 300 kilowatts of air conditioning load are moved to 17 to 19 hours, realizing peak clipping operation. At the same time, the photovoltaic array orientation is adjusted to match the power generation peak with the electricity peak.

[0027] In step S104, the electricity plan data of the industry-university-research area is obtained, the electricity load change trend of the industry-university-research area is analyzed, and the time period of peak clipping electricity is determined. According to the characteristics of the electricity load of the teaching area rising during the school period, the electricity plan of the industry-university-research area is adjusted to support the electricity demand of the teaching area.

[0028] The load power value is fitted and calculated by using a long short-term memory network, and a power consumption feature data table containing a time stamp and a load value is calculated through a sliding time window; a Gaussian kernel similarity matrix is constructed by using a spectral clustering algorithm according to the power consumption feature data table, and a load period distribution data table is calculated through a Pearson correlation coefficient; a power load growth prediction result is calculated through a recurrent neural network for the load period distribution data table; a load balancing constraint condition is set according to the power load growth prediction result, and a power plan adjustment scheme containing a device number and an adjustment time period is calculated through an integer programming method.

[0029] Specifically, the equipment operation schedule, load power value, and start-stop state record are obtained from the production-teaching-research area electricity plan database. The four-layer long short-term memory network is used to fit the load variation trend, and the number of hidden layer neurons is set to 128. The electricity characteristic data table containing time stamp, load value, and change rate is calculated through a 12-hour sliding time window. According to the electricity characteristic data table, the spectral clustering algorithm is used to decompose the load curve, and the number of clusters is set to 3. The similarity matrix is constructed using the Gaussian kernel function, and the time period distribution data table of each type of load is calculated by the Pearson correlation coefficient to generate the time period division result of peak-shaving electricity. The total electricity consumption, peak-valley electricity difference, and daily average electricity consumption fluctuation data are collected from the teaching area electricity monitoring unit to extract the load characteristics during the school season. The three-layer recurrent neural network is used to predict the future electricity load growth trend, and the time step is set to 1 hour. According to the load growth prediction result, the load balance constraint condition is set, including the maximum load limit value of a single period, the load transfer time interval, and the equipment start-stop frequency limit to generate the load adjustment constraint table. For the load adjustment constraint table and the peak-shaving electricity time period division result, the integer programming method is used to rearrange the production-teaching-research area equipment operation time to obtain the electricity plan adjustment scheme containing equipment number, adjustment period, and load change value. According to the electricity plan adjustment scheme, the dynamic programming algorithm is used to calculate the load transfer path to generate the load adjustment execution table containing start and end time, transfer load value, and equipment operation state. The production-teaching-research area electricity plan data collection covers the operation parameters of various types of electricity-consuming equipment. The large experimental equipment operates from 9 am to 17 pm on weekdays, and the power of a single device is 50 kW, which needs to be continuously operated for 4 hours after starting. The four-layer long short-term memory network is used to fit the load variation trend, the input layer receives 24 hours of historical load data, the number of hidden layer neurons is 128, and the load variation characteristics are calculated through a 12-hour sliding window. The experiment shows that the average error rate of this network structure in load prediction is less than 8%. The spectral clustering method is used for load curve decomposition, and the electricity load is divided into three categories: basic load, fluctuation load, and peak load. The basic load mainly comes from lighting and air conditioning systems, accounting for 35% of the total load, the fluctuation load includes office equipment and small experimental equipment, accounting for 40%, and the peak load is generated by large experimental equipment, accounting for 25%. Through Pearson correlation coefficient analysis, it is found that the peak load and working days show a significant positive correlation, with a correlation coefficient of 0.85. The electricity load of the teaching area in the school season shows a significant upward trend, with the total electricity consumption increasing from 2000 kWh per day during the holiday to 4500 kWh per day, and the peak-valley difference expanding from 800 kWh to 2000 kWh. The three-layer recurrent neural network is used to predict the future load growth trend, and the time step is set to 1 hour. The prediction result shows that the load growth rate is highest in the first week after school, reaching 50%. The load adjustment constraint condition contains multiple dimensions, and the maximum load limit value of a single period is set to 3000 kWh to avoid overloading of power equipment.The device start-stop interval is not less than 2 hours, limiting frequent on-off. The daily start-stop times of large experimental equipment are not more than 2 times, ensuring the continuity of the experiment. Under these constraints, the optimal scheduling scheme is calculated by integer programming method. The calculation of load transfer path uses dynamic programming algorithm, which divides 24 hours into 48 time periods. Taking a working day as an example, the running time of 4 large experimental equipment is delayed from 14 o'clock to 17 o'clock, and the cumulative transferred load is 200 kilowatt-hours. At the same time, the start-up time of 10 medium-sized experimental equipment is adjusted from 9 o'clock to 8 o'clock, and the transferred load is 150 kilowatt-hours. Through load peak shifting, the original 95% device simultaneous operation rate is reduced to 75%. When executing the load adjustment scheme, the transfer is carried out according to the priority order of the equipment. The air conditioning system is adjusted to the energy-saving temperature, the lighting system is automatically dimmed according to the natural light intensity, and the office equipment adopts the time-sharing opening strategy. Experimental equipment is sorted according to the length and urgency of the experiment, and the running time of teaching experimental equipment is guaranteed first, and research experimental equipment is appropriately delayed. Through fine load management, the dynamic balance of electricity load between teaching area and industry-university-research area is realized.

[0030] In step S105, based on the electricity plan data of the campus and the industry-university-research area, the deployment of the photovoltaic array dynamic strategy is carried out, the target deployment position of the photovoltaic array in different time periods is calculated combined with the moving ability and power generation efficiency of the photovoltaic array, and the dynamic deployment scheme of the photovoltaic array is generated.

[0031] The load curve record is obtained from the electricity plan database, the long short-term memory network is used for periodic decomposition of the load curve record, the photovoltaic power generation potential distribution map is obtained according to the light intensity data and the solar panel conversion efficiency curve; for the photovoltaic power generation potential distribution map, the deep learning segmentation algorithm is used to extract the deployable area boundary point and the shadow projection range, and the power generation efficiency attenuation coefficient is obtained by cosine similarity calculation; according to the power generation efficiency attenuation coefficient, the equipment motion parameters are obtained from the mobile device controller, and the equipment movement constraint table is generated through the site access restriction; for the equipment movement constraint table, the particle swarm optimization algorithm is used to optimize the deployment position, and the moving time table is generated through the dynamic time planning algorithm.

[0032] Specifically, the load curve records of the campus and the industry-university-research area are obtained from the electricity plan database. A three-layer long short-term memory network is used to periodically decompose the electricity load data, with 256 hidden neurons. The time series data of light intensity and the solar panel conversion efficiency curve are combined to obtain the photovoltaic power generation potential distribution map of different time periods through two-dimensional convolution operation. According to the photovoltaic power generation potential distribution map and the three-dimensional scanning data of the site, the deep learning segmentation algorithm is used to extract the spatial boundary points and the building shadow projection range of the deployable area. The cosine similarity is used to calculate the building shading loss, and the power generation efficiency attenuation coefficient of each area is obtained. The equipment movement parameters, including maximum moving speed, turning radius, and load capacity, are obtained from the mobile device controller, and the equipment movement constraint table is generated combined with the site traffic restrictions. For the power generation efficiency attenuation coefficient and the equipment movement constraint table, the particle swarm optimization algorithm is used to optimize the deployment location, with a population size of 200 and an iteration number of 500. The deployment planning table containing the target location coordinates and the equipment orientation angle is obtained. The Euclidean distance between adjacent deployment locations is calculated according to the deployment planning table, and the dynamic time planning algorithm is used to arrange the equipment movement time sequence, generating a movement time table containing the start and end times and the equipment number. The ant colony algorithm is used to plan the equipment moving path, with a pheromone decay coefficient of 0.3 and a local search probability of 0.1. The equipment relocation route map is generated through path smoothing processing. The electricity load data is processed using a three-layer long short-term memory network, with the input layer receiving 24 hours of historical data, 256 hidden neurons, and the output layer predicting the load change in the next 12 hours. The network training data includes electricity records for the past year, with a clear double-peak feature on weekdays. The first peak is from 9 am to 11 am, and the second peak is from 2 pm to 4 pm, with a peak load of 4000 kilowatt-hours. The three-dimensional scanning data of the site reflects the shading effect of buildings on photovoltaic power generation. For example, the east side of the teaching building is shaded from 8 am to 10 am, and the power generation efficiency attenuation coefficient for this period is 0.6, while the attenuation coefficient at noon is only 0.1. The parking area is less affected by shading throughout the day and is suitable for deploying photovoltaic arrays. The mobile device of the photovoltaic array adopts a track design, with a maximum moving speed of 0.5 meters per second, a turning radius not less than 4 meters, and a single maximum load of 500 kilograms. The site traffic restrictions include a road slope not exceeding 15 degrees, a passage width not less than 3 meters, and a safety distance of at least 1 meter between the equipment during transfer and fixed facilities. The particle swarm optimization algorithm is used for deployment location optimization, with a population size of 200 and an iteration number of 500. The optimization objectives include maximizing power generation efficiency and minimizing moving distance. The optimization results for a typical working day show that the photovoltaic array is deployed in the east area of the parking lot from 8 am to 11 am, moved to the central area of the parking lot from 11 am to 2 pm, and deployed in the west area of the parking lot from 2 pm to 5 pm, achieving the best power generation effect throughout the day.The device movement timing arrangement considers the shadow change rule, and calculates the optimal movement time through a dynamic time planning algorithm. Taking the parking area as an example, the best time for moving from the east area to the middle area is 10:30, at which time the shadow boundary just moves out of the east area, and the device transfer does not affect the power generation efficiency. Similarly, the best time for moving from the middle area to the west area is 13:30. The movement path planning adopts an improved ant colony algorithm, which introduces a local search mechanism on the basis of the standard ant colony algorithm, sets the pheromone decay coefficient to 0.3, and the local search probability to 0.1. The path smoothing processing is performed through Bezier curve fitting, and the path nodes with a turning radius less than 4 meters are optimized to generate a smooth route that meets the device turning constraint. In actual deployment, the route length for moving the device from the east area to the middle area of the parking lot is 50 meters, containing 2 turning nodes, and the moving time is 3 minutes. The minimum turning radius after path smoothing processing is 4.5 meters, which meets the device motion performance requirements.

[0033] In step S106, the dynamic deployment strategy scheme of the photovoltaic array is compared with the power consumption plan data to determine whether the photovoltaic power generation capacity meets the power consumption demand of the campus and the industry-university-research area. If yes, the deployment scheme is executed; if not, the deployment position of the photovoltaic array is adjusted or the peak-shaving production mechanism is started.

[0034] The movement timing table and position coordinates are obtained according to the photovoltaic array deployment scheme database, the predicted power generation data table of each deployment position is obtained through radial basis kernel function support vector regression, the predicted power generation data table contains timestamp, predicted power generation and shading rate; the multi-layer perceptron is used to decompose the power consumption load data, the multi-layer perceptron receives temperature, humidity and timestamp input features to obtain the power consumption demand data table of different areas at each time period; numerical comparison is performed on the predicted power generation data table and the power consumption demand data table to obtain a supply-demand balance table containing time period, gap value and area identifier; if the gap value in the supply-demand balance table exceeds the total power consumption threshold, the integer programming method is used to optimize the deployment position of the photovoltaic array, and the device scheduling instruction set containing target coordinates, movement path and execution time is obtained through optimization.

[0035] Specifically, the mobile timing table and position coordinates are obtained from the photovoltaic array deployment scheme database, the radial basis kernel function support vector regression is used to predict the power generation of each deployment position, the penalty coefficient is set to 1.0, the light resource distribution and shadow blocking loss of each time period are calculated through a 4-hour sliding time window, and a power supply data table containing time stamp, predicted power generation and blocking rate is generated. According to the power consumption plan of the campus and the industry-university-research area, a four-layer multilayer perceptron is used to decompose the power consumption load data, the number of hidden layer neurons is 128, the input features include temperature, humidity and time stamp, and the power consumption demand data table of different areas at each time period is obtained through time series analysis. According to the power supply data table and the power consumption demand data table, the power supply gap value of each period is calculated by numerical comparison, and a supply-demand balance table containing time period, gap value and area identifier is generated. For the supply-demand balance table data, the power supply gap threshold is set to 20% of the total power consumption, and if the gap value exceeds the threshold, the integer programming method is used to optimize the photovoltaic array deployment position, and the constraint conditions include site area limit, equipment turning radius and passing space requirement. According to the deployment position optimization result, the device scheduling instruction set containing target coordinates, moving path and execution time is generated. The dynamic programming algorithm is used to calculate the load adjustment scheme of each area, and the maximum load transfer amount of each period is set to 30% of the total load, and the peak shifting operation scheme containing device number, peak shifting period and load adjustment amount is generated. The support vector regression method is used for photovoltaic array power generation prediction, the radial basis kernel function is used to process nonlinear characteristics, and the penalty coefficient is set to 1.0 to balance the model complexity. Under typical sunny weather conditions, the power generation of a certain photovoltaic panel group from 8 to 16 shows a parabolic distribution, reaching a peak of 280 watt-hours at 12 noon, while shadow blocking causes a 20% to 40% reduction in power generation efficiency. A 4-hour sliding time window is used for calculation to accurately capture the dynamic changes of light intensity and blocking loss. Four-layer multilayer perceptron is used for power load decomposition, and temperature, humidity and time stamp data are received in the input layer, with 128 hidden layer neurons. In practical application, the daily power consumption of a certain teaching building is 2400 kilowatt-hours, of which air conditioning load accounts for 45%, lighting load accounts for 25%, and power load accounts for 30%. For every 1 degree increase in temperature, the air conditioning load increases by 5%, and for every 10% increase in humidity, the air conditioning load increases by 3%. The supply-demand balance analysis calculates the power supply gap in units of hours, the first power consumption peak is from 9 to 11 on weekdays, the photovoltaic power generation is 800 kilowatt-hours, the power consumption demand is 1200 kilowatt-hours, and the power supply gap is 400 kilowatt-hours, which exceeds the threshold of 20% of the total power consumption. From 14 to 16, the second peak, the photovoltaic power generation decreases to 600 kilowatt-hours, the power consumption demand reaches 1400 kilowatt-hours, and the power supply gap further expands. The photovoltaic array deployment position optimization considers multiple constraint conditions, the site area limit is 2000 square meters, the equipment turning radius is not less than 4 meters, and the passing channel width is not less than 3 meters.The optimization results show that moving the rooftop photovoltaic array to the parking lot area during the 14-16 period with the largest power supply gap increases power generation by 30%, but still cannot fully meet the electricity demand. The device scheduling instruction plan specifies the path points, with a relocation distance of 150 meters from the teaching building roof to the parking lot, including 3 turning nodes, a device moving speed of 0.5 meters / second, and an estimated completion time of 6 minutes. Considering the turning radius limit, a Bezier curve is used for trajectory smoothing at each turning point to ensure smooth operation of the device. The load adjustment scheme calculates the optimal scheduling sequence through dynamic programming, and the maximum load transfer amount in a single period does not exceed 30% of the total load. For the 14-16 period with the largest power supply gap, 300 kW of non-critical experimental equipment load is moved to the 16-18 period, and the air conditioning temperature set value is adjusted to reduce 150 kW of air conditioning load, achieving dynamic balance between supply and demand, and the power supply gap in each period is controlled within the threshold range.

[0036] In step S107, according to the actual execution data of the photovoltaic array dynamic deployment scheme, the comparison results of photovoltaic power generation and electricity demand are obtained, the effect of power coordination and mutual aid is analyzed, the photovoltaic array deployment strategy and peak shifting production mechanism are optimized, and a school-enterprise joint power supply scheme is generated.

[0037] The array power generation data, movement path records and running state information are obtained from the photovoltaic monitoring unit, a long short-term memory network is used for analysis, and a supply and demand data table containing time stamps, matching rates and difference values is obtained; according to the supply and demand data table, a back propagation neural network is used to analyze the correlation between photovoltaic array deployment location and power generation efficiency, and a location evaluation table containing region number, power generation coefficient and time period distribution is obtained; for the location evaluation table, an integer programming method is used to optimize the photovoltaic array layout, the optimization calculation is based on site bearing limit, equipment spacing requirement and passing width constraint, and a deployment adjustment scheme containing device number, target location and execution time sequence is obtained; according to the deployment adjustment scheme, an ant colony algorithm is used to calculate the shortest path of device migration, the ant colony algorithm is based on pheromone decay coefficient and local search probability for path optimization, and a device migration instruction set containing path node, turning angle and moving time length is obtained.

[0038] Specifically, the power generation data, moving path records, and equipment operating status during the dynamic deployment of the array are obtained from the photovoltaic monitoring unit. A three-layer long short-term memory network is used to analyze the supply-demand matching time series data, with 256 hidden neurons. The supply-demand ratio and peak-valley difference are calculated through a 4-hour sliding time window to generate a supply-demand data table containing timestamps, matching rates, and difference values. Based on the supply-demand data table, a four-layer backpropagation neural network is used for correlation analysis of the photovoltaic array deployment location and power generation efficiency. The input layer features include light intensity, shadow blocking rate, and slope angle. The K-means clustering algorithm is used to divide the site into high, medium, and low power generation efficiency areas, generating a location evaluation table containing region numbers, power generation coefficients, and time period distribution. Based on the location evaluation table, site load-bearing restrictions, equipment spacing requirements, and passage width constraints are set. Integer programming is used to optimize the photovoltaic array layout, generating a deployment adjustment scheme containing equipment numbers, target locations, and execution sequences. The deployment adjustment scheme execution path is planned, and the improved ant colony algorithm is used to calculate the shortest path with a pheromone decay coefficient of 0.3 and a local search probability of 0.1, generating a device migration instruction set containing path nodes, turning angles, and moving time lengths. Based on the device migration instruction set, the dynamic programming algorithm is used to divide the peak-shaving production period, with a single period maximum load transfer amount set to 30% of the total load, generating a peak-shaving operation table containing start and end times and load adjustment amounts. The hierarchical clustering method is used to classify regional electricity loads with a clustering distance threshold of 20% of the total load, and the peak-shaving operation table is combined to generate a school-enterprise joint power supply scheme containing region numbers, operation periods, and load distribution. The supply-demand matching analysis of the photovoltaic array is processed using a three-layer long short-term memory network, with the input layer receiving 24-hour historical data and 256 hidden neurons. Through a 4-hour sliding window, the supply-demand dynamic changes are calculated. Taking a working day as an example, the supply-demand matching rate from 9 am to 11 am is 75%, the electricity demand is 1200 kWh, the photovoltaic power generation is 900 kWh, and the peak-valley difference is 300 kWh. From 12 pm to 14 pm, the matching rate increases to 90%, mainly due to the peak light intensity. The deployment location evaluation uses a four-layer backpropagation neural network with input features including light intensity, shadow blocking rate, and slope angle. The site is divided into high, medium, and low power generation efficiency areas through K-means clustering. The parking lot area belongs to the high-efficiency power generation area with a power generation coefficient of 0.95, suitable for deploying photovoltaic arrays from 9 am to 16 pm. The east side of the teaching building is affected by building shading, with a power generation coefficient of only 0.6, making it unsuitable for deployment. The site constraint conditions include multiple dimensions, such as a roof load-bearing limit of 60 kg per square meter, a device spacing requirement of not less than 1.5 meters, and a passage width of not less than 3 meters. The optimization results show that the parking lot area is 2000 square meters, with an actual available area of 1400 square meters considering various constraints. Arranging 120 photovoltaic arrays can achieve an annual average power generation of 280,000 kWh.The device moving path is planned by using the ant colony algorithm, the pheromone decay coefficient 0.3 is used to control the convergence speed of the path, and the local search probability 0.1 is used to enhance the path diversity. The migration route from the roof of the teaching building to the parking lot is 150 meters long, containing 3 turning nodes, and the turning angle of each node is not more than 60 degrees. The moving process is expected to take 6 minutes. The off-peak production period is divided by using the dynamic programming method, and 24 hours is divided into 48 time periods, each period being 30 minutes. The maximum load transfer amount of a single period is set to 30% of the total load, which is equivalent to 900 kilowatt-hours. The load of large equipment in the experimental area is transferred from 14:00 to 16:00 to 16:00 to 18:00, and the cumulative transferred load is 600 kilowatt-hours, accounting for 20% of the total load of the period. Hierarchical clustering divides the electrical equipment into three levels, and the clustering distance threshold is 20% of the total load. Among them, the lighting and air conditioning load is the basic load, accounting for 45%, and the running time is fixed; the office and teaching equipment is the medium load, accounting for 30%, and the running time can be adjusted appropriately; the scientific research and experimental equipment is the high load, accounting for 25%, and the running time can be arranged flexibly according to the power supply situation. In the school-enterprise joint power supply scheme, the power consumption peak of the teaching area is arranged from 9:00 to 12:00, and the power consumption peak of the production-study-research area is arranged from 13:00 to 16:00, realizing the off-peak mutual aid.

[0039] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit the scope of protection of the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the essence and scope of the technical solutions of the present application.

Claims

1. A new energy power intelligent allocation method, characterized in that, The method comprises: According to the historical electricity consumption data of the campus and the industry-university-research area, the electricity consumption load change trend of the teaching area and the industry-university-research area is extracted, the electricity consumption cycle difference between the holiday and the school season is analyzed, and the electricity consumption complementary time period is determined; Obtain the power generation efficiency data of the campus photovoltaic array, combine the campus space layout information and the electricity consumption complementary time period, update the photovoltaic array deployment scheme, and calculate the updated power generation potential, determine the characteristics of the electricity consumption load increase of the teaching area, and determine the target position of the photovoltaic array movement; According to the target position of the photovoltaic array movement, the power generation data after the photovoltaic array moves is calculated, and the electricity consumption demand data of the industry-university-research area is combined to determine whether the photovoltaic power generation meets the electricity consumption demand of the industry-university-research area, if not, the peak load shifting production mechanism is started to generate the photovoltaic array deployment position in the electricity consumption load reduction stage of the teaching area; Obtain the electricity consumption plan data of the industry-university-research area, analyze the electricity consumption load change trend of the industry-university-research area and determine the time period of peak load shifting, according to the characteristics of the electricity consumption load increase of the teaching area during the school period, adjust the electricity consumption plan of the industry-university-research area to support the electricity consumption demand of the teaching area; Based on the electricity consumption plan data of the campus and the industry-university-research area, the dynamic strategy of the photovoltaic array is deployed, the target deployment position of the photovoltaic array in different time periods is calculated by combining the moving ability and the power generation efficiency of the photovoltaic array, and the dynamic deployment scheme of the photovoltaic array is generated; By comparing the dynamic deployment strategy scheme of the photovoltaic array with the electricity consumption plan data of the campus and the industry-university-research area, it is determined whether the photovoltaic power generation meets the electricity consumption demand of the campus and the industry-university-research area, if it meets, the deployment scheme is executed, if it does not meet, the deployment position of the photovoltaic array is adjusted or the peak load shifting production mechanism is started.

2. The method of claim 1, wherein, According to the historical electricity consumption data of the campus and the industry-university-research area, the electricity consumption load change trend of the teaching area and the industry-university-research area is extracted, the electricity consumption cycle difference between the holiday and the school season is analyzed, and the electricity consumption complementary time period is determined, comprising: Obtain the electricity consumption load curve collected by the online monitoring sensor, repair the missing data points in the electricity consumption load curve by using the moving average interpolation method, obtain the electricity consumption load distribution data table containing time stamp, load value and peak valley identification; According to the electricity consumption load distribution data table, set the electricity consumption load change rate threshold, mark the data points exceeding the threshold, and obtain the cycle characteristic table containing time period, load change amount and duration; For the cycle characteristic table, the electricity consumption load mean, standard deviation and coefficient of variation are calculated by using hierarchical clustering method, and the interval boundary time point sequence is obtained; According to the interval boundary time point sequence, the sliding time window method is used for segmented processing of the equipment operation characteristic data, the time dimension of each area electricity consumption load is redistributed by using the dynamic programming algorithm, and the complementary interval load regulation scheme is obtained.

3. The method of claim 1, wherein, The electricity consumption plan data of the campus and the industry-university-research area is obtained, the electricity consumption load change trend of the industry-university-research area is analyzed, and the time period of peak load shifting is determined, according to the characteristics of the electricity consumption load increase of the teaching area during the school period, the electricity consumption plan of the industry-university-research area is adjusted to support the electricity consumption demand of the teaching area, comprising: The array power generation data and the light intensity data collected by the photovoltaic monitor are acquired, the light resource distribution data table is obtained by Fourier transform on the light intensity data, and the light resource distribution data table contains timestamp, light intensity and power generation efficiency; According to the light resource distribution data table and the three-dimensional laser scanning data, the site boundary coordinate point set is extracted by using a deep learning segmentation algorithm, and the site layout constraint table is calculated by vector projection; According to the site layout constraint table and the power load curve record, the power load curve record is calculated by using a sliding window, and the power feature data table is obtained; According to the power feature data table and the site layout constraint table, the solar panel arrangement is optimized by using a multi-objective genetic algorithm, and the photovoltaic array deployment scheme is obtained, which contains equipment number, installation position and orientation angle.

4. The method of claim 1, wherein, According to the target position of the photovoltaic array movement, the power generation data after the movement of the photovoltaic array is calculated, and the power demand data of the production-research-education zone is combined to determine whether the photovoltaic power generation meets the power demand of the production-research-education zone, if not, the peak-shaving production mechanism is started to generate the photovoltaic array deployment position in the power load reduction stage of the teaching area, including: The light intensity is calculated by using Gaussian process regression according to the spatial coordinate data of the photovoltaic array, the kernel function of the Gaussian process regression is selected as the radial basis function, and the photovoltaic panel power generation prediction data table is obtained; The building power load is decomposed by using a random forest algorithm, the feature dimension of the random forest algorithm contains timestamp, temperature and passenger flow data, and the sub-item power data table is obtained; The proportion of each type of load is calculated by using a hierarchical clustering method on the sub-item power data table, the hierarchical clustering method sets the clustering distance threshold as the total load percentage parameter, and the load classification result table is obtained; According to the power generation prediction data table and the load classification result table, the power supply gap value is calculated by using a time sequence matching method, if the power supply gap value is greater than a preset threshold, the peak-shaving scheduling instruction set is generated.

5. The method of claim 1, wherein, The power consumption plan data of the production-research-education zone is acquired, the power load change trend of the production-research-education zone is analyzed and the time period of peak-shaving power consumption is determined, according to the characteristics of the power load increase of the teaching area during the school opening period, the power consumption plan of the production-research-education zone is adjusted to support the power demand of the teaching area, including: The load power value is fitted by using a long short-term memory network, and the power feature data table containing timestamp and load value is calculated by using a sliding time window; According to the power feature data table, a Gaussian kernel similarity matrix is constructed by using a spectral clustering algorithm, and a load time period distribution data table is calculated by using a Pearson correlation coefficient; According to the load time period distribution data table, the power load growth prediction result is calculated by using a recurrent neural network; According to the power load growth prediction result, the load balancing constraint condition is set, and the power consumption plan adjustment scheme containing equipment number and adjustment time period is calculated by using an integer programming method.

6. The method of claim 1, wherein, Based on the electricity consumption plan data of the campus and industry-university-research area, a dynamic strategy for deploying photovoltaic arrays is implemented. Combining the mobility and power generation efficiency of the photovoltaic arrays, the target deployment locations of the photovoltaic arrays are calculated for different time periods, generating a dynamic deployment scheme for the photovoltaic arrays, including: Load curve records are obtained from the electricity consumption plan database. The long short-term memory network is used to decompose the load curve records periodically. A photovoltaic power generation potential distribution map is obtained based on the light intensity data and the solar panel conversion efficiency curve. For the photovoltaic power generation potential distribution map, a deep learning segmentation algorithm is used to extract the boundary points of the deployable area and the shadow projection range, and the power generation efficiency attenuation coefficient is calculated by cosine similarity. Based on the power generation efficiency attenuation coefficient, the equipment motion parameters are obtained from the mobile device controller, and a device motion constraint table is generated by the site access restrictions. For the aforementioned equipment movement constraint table, a particle swarm optimization algorithm is used to optimize the deployment location, and a movement schedule is generated through a dynamic time planning algorithm.

7. The method of claim 1, wherein, The process involves comparing the dynamic deployment strategy of the photovoltaic array with the electricity consumption plan data of the campus and the industry-university-research area to determine whether the photovoltaic power generation meets the electricity demand of the campus and the industry-university-research area. If it does, the deployment plan is executed; if not, the deployment location of the photovoltaic array is adjusted or a peak-shaving production mechanism is activated, including: The movement time series table and location coordinates are obtained from the photovoltaic array deployment scheme database. The predicted power generation data table for each deployment location is obtained through radial basis function support vector regression. The predicted power generation data table includes timestamp, predicted power generation and shading rate. A multilayer sensor is used to decompose the power load data. The multilayer sensor receives temperature, humidity and timestamp input features to obtain a table of power demand data for different areas at different time periods. A supply and demand balance table is obtained by comparing the predicted power generation data table with the electricity demand data table, which includes time period, gap value and regional identifier. If the gap value in the supply and demand balance table exceeds the total electricity consumption threshold, then an integer programming method is used to optimize the deployment location of the photovoltaic array, and the optimization yields a set of equipment scheduling instructions that includes target coordinates, movement paths, and execution times.

8. The method of claim 1, wherein, The method further includes: obtaining the comparison results of photovoltaic power generation and electricity demand based on the actual execution data of the photovoltaic array dynamic deployment scheme, analyzing the effect of power synergy and mutual assistance, optimizing the photovoltaic array deployment strategy and peak-shifting production mechanism, and generating a university-enterprise joint power supply scheme.

9. The method of claim 8, wherein, The process involves obtaining a comparison between photovoltaic power generation and electricity demand based on actual execution data from the dynamic deployment plan of the photovoltaic array, analyzing the effect of power synergy and mutual assistance, optimizing the photovoltaic array deployment strategy and peak-shaving production mechanism, and generating a university-enterprise collaborative power supply plan, including: Data on array power generation, movement path records, and operational status information are obtained from the photovoltaic monitoring unit. Long Short-Time Memory (LSTM) network is used for analysis to obtain a supply and demand data table containing timestamps, matching rates, and difference values. According to the supply-demand data table, a back propagation neural network is used to analyze the correlation between the photovoltaic array deployment location and the power generation efficiency, divide the site power generation efficiency area, and obtain a location evaluation table containing area number, power generation coefficient and time period distribution; For the location evaluation table, an integer programming method is used to perform optimization calculation on the photovoltaic array layout, the optimization calculation is based on site load bearing limit, equipment spacing requirement and passing width constraint, and a deployment adjustment scheme containing equipment number, target location and execution time sequence is obtained; According to the deployment adjustment scheme, an ant colony algorithm is used to calculate the shortest path of equipment migration, the ant colony algorithm is based on pheromone decay coefficient and local search probability to optimize the path, and an equipment migration instruction set containing path node, turning angle and moving time length is obtained.

Images