Multivariate time sequence prediction method and system for waste plastic output

Abstract

The invention belongs to the technical field of waste plastic prediction, and discloses a waste plastic generation quantity multivariate time sequence prediction method and system, and the method comprises the steps: dividing a predicted waste plastic generation region into an environmental sanitation system coverage region and an environmental sanitation system uncoverage region; the plastic garbage in the coverage area of the environmental sanitation system passes through a garbage transfer station or a garbage recycling facility, and the non-coverage area of the environmental sanitation system is averagely divided into # imgabs0 grids; the method comprises the following steps: dividing inbound garbage of a garbage transfer station into recoverable garbage and unrecoverable garbage, and collecting the daily unrecoverable amount of a plastic part in the unrecoverable garbage; collecting the recyclable amount of daily plastic garbage of the garbage recycling facility; collecting actual measurement of daily plastic garbage of each grid in a selected grid set in an uncovered area of the environmental sanitation system; calculating the calculated amount of daily plastic garbage in each grid of the uncovered area of the environmental sanitation system based on actual measurement; and the prediction effect of the prediction model can be obviously improved.

Description

Technical Field

[0001] The present invention relates to the field of waste plastic prediction technology, and more specifically, to a multivariate time series prediction method and system for waste plastic generation. Background Art

[0002] Predicting the amount of waste plastic generated is an important part of assessing plastic pollution, formulating recycling policies and optimizing resource management; it can provide strong support for waste disposal planning, environmental protection decision-making, public education and technological innovation.

[0003] In the prior art, the calculation methods of waste plastic generation include per capita statistics, life cycle analysis and material flow analysis. However, there are significant deficiencies when the floating population and the products containing plastics consumed frequently enter and exit the preset area, which are mainly manifested in the following aspects: Random discarding, changes in floating population, plastic waste with water and direct weighing of residues lead to unstable data collection, making it impossible to track the true level of waste plastic generation. It is impossible to provide a reliable reference standard value for the prediction model, affecting the credibility of the prediction results.

[0004] Therefore, existing technologies make it difficult to accurately predict the amount of waste plastic generated in areas with frequent population mobility, and there is an urgent need for a prediction method with accurate data and strong applicability.

[0005] In view of this, the present invention proposes a multivariate time series prediction method and system for waste plastic generation to solve the above problems. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art and achieve the above-mentioned purpose, the present invention provides the following technical solution: a multivariate time series prediction method for waste plastic generation, comprising: The predicted waste plastic generation area is divided into areas covered by the sanitation system and areas not covered by the sanitation system; the plastic waste in the area covered by the sanitation system passes through the garbage transfer station or garbage recycling facilities, and the area not covered by the sanitation system is evenly divided into A grid; Separate the incoming garbage at the garbage transfer station into recyclable and non-recyclable garbage, and collect the daily non-recyclable amount of plastic in the non-recyclable garbage; Collect the recyclable amount of plastic waste from recycling facilities every day; Collect daily measurements of plastic waste for each grid in a selected set of grids in areas not covered by sanitation systems; Calculate the amount of daily plastic waste in each grid in areas not covered by the sanitation system based on actual measurements; Calculate the daily disposal volume in areas covered by the sanitation system based on the non-recyclable volume and the recyclable volume; calculate the daily pollution volume in areas not covered by the sanitation system based on the calculated volume; calculate the daily waste plastic generation volume based on the disposal volume and the pollution volume; Based on preset Set up training data and train the prediction model to predict the future The amount of waste plastic produced per day, is the length of the time window; the training data includes non-recyclable amount, recyclable amount, pollution amount, and waste plastic generation amount; Using the time series data of non-recyclable, recyclable and polluted quantities through the prediction model, the future The amount of waste plastic generated per day.

[0007] Furthermore, the calculation amount includes the calculation amount of each grid in the selected grid set and the calculation amount of each grid in the unselected grid set. Day selected grid collection The computational effort for each grid is equal to the actual measurement of the day; the computational effort for each grid in the unselected grid set is calculated based on the actual measurement of the selected grid found by the minimum encirclement strategy.

[0008] Further, the method for calculating the amount of each grid in the unselected grid set includes: At the time, The geometric centroid of the unselected mesh is the center point. The average value of the actual measurements of all selected grids on the first day found using the minimum encirclement strategy is taken as the The amount of calculation for the unselected grids on the first day; At the time, The geometric centroid of the unselected grids is taken as the center point, and the increment of the actual measurement of each selected grid on the second day relative to the first day found by the minimum encirclement strategy is calculated. The average of all increments is taken and recorded as the neighborhood increment. The calculation amount of the unselected grid is added to the neighborhood increment to get the second day The amount of calculation for unselected grids; At the time, The geometric centroid of the unselected mesh is used as the center point, and the minimum enclosing circle strategy is used to calculate the The increment of the actual measurement of the day relative to the previous day is taken, and the average value of the increment is recorded as the neighborhood increment; Unselected grids in The calculation amount of the day is weighted averaged with the increment of the previous day and the increment of the neighborhood according to the preset weights to get the Unselected grids in The increment of the calculation amount of the day relative to the previous day is recorded as the recent increment; Unselected grids in The calculation amount of the day is added to the recent increment to get the Unselected grids Days of calculations.

[0009] Furthermore, the minimum encirclement strategy includes: Step 1: Set a radius, take the geometric centroid of the currently unselected grid as the center point, and gradually increase the radius by multiples to form a circular encirclement; Step 2, start searching from the encirclement with the minimum multiple radius. If at least one selected grid with a geometric centroid falling within the encirclement is found, record the selected grid falling within the encirclement, and then stop searching; Step 3: If no selected grid with a geometric center of mass falling within the encirclement is found, the radius of the encirclement is increased by 1 times based on the current radius, and step 2 is repeated until at least one selected grid with a geometric center of mass falling within the encirclement is found.

[0010] Furthermore, based on the preset and Methods for setting up training data include: The first Day to Day All feature data of the day are taken as input data, and the The daily waste plastic generation is used as the output data, and the input data and output data are combined into a set of training data. The daily feature data includes the non-recyclable amount, recyclable amount, pollution amount, and waste plastic generation.

[0011] Furthermore, the area covered by the sanitation system is the area in the preset area where regular cleaning and regular collection and transportation of garbage are arranged; the area not covered by the sanitation system is the area in the preset area where regular cleaning or regular collection and transportation of garbage is not arranged; the outbound garbage of the garbage transfer station in the area covered by the sanitation system is sent to the garbage treatment facilities and garbage recycling facilities respectively; the inbound garbage of the garbage treatment facilities comes only from the garbage transfer stations and garbage recycling facilities; the inbound garbage of the garbage recycling facilities comes from various recycling channels, including recyclable garbage sent from the garbage transfer stations; the outbound garbage of the garbage recycling facilities is sent to the garbage treatment facilities, which is the remaining residue after the recycling process.

[0012] Furthermore, the method for collecting the non-recyclable amount includes: In the Every day, at each garbage transfer station, the RFID tag reading device installed on the incoming garbage conveyor belt reads the RFID tag of each piece of garbage passing through; the incoming garbage is separated into recyclable garbage and non-recyclable garbage according to the recyclable information on the RFID tag; the mass of plastic garbage is determined according to the RFID tag, and the mass of the plastic part of the non-recyclable garbage after separation is calculated; the mass of the plastic part of the non-recyclable garbage in all garbage transfer stations is added up to get the mass of the plastic part of the non-recyclable garbage in the first garbage transfer station. The non-recyclable amount of the day.

[0013] Furthermore, the method for collecting the recyclable amount includes: In the Every day, at each waste recycling facility, an RFID tag reading device installed on the inbound recyclable waste conveyor belt reads the RFID tag of each piece of waste passing through; the mass of plastic waste is determined based on the RFID tag, and the plastic portion of the inbound recyclable waste is counted; the mass of the plastic portion of the inbound recyclable waste of all waste recycling facilities is added up to obtain the mass of the plastic portion of the inbound recyclable waste of all waste recycling facilities. The amount of recyclables per day.

[0014] The method for collecting the actual measurement of each grid in the selected grid set includes: In the Day, randomly selected The selected mesh collection consists of meshes. ; For each grid in the selected grid set, a drone carrying an RFID tag reader is used to conduct a complete search along a preset path in the land part, and an unmanned submarine carrying an RFID tag reader is used to conduct a complete search along a preset path in the water part; the read RFID tags are recorded, the mass of the plastic waste is determined based on the RFID tags, the mass of the plastic waste in the grid is accumulated, and the actual measurement of each grid in the selected grid set is obtained.

[0015] Furthermore, the calculation method of the appropriate amount includes: The first The non-recyclable amount of the first day is Add up the recyclable amount of the day to get The proper measure of heaven; The method for calculating the pollution amount includes: In the sky, , for areas not covered by the sanitation system The recent increments of the computational amount of each grid are summed to obtain The amount of pollution per day; The calculation method of waste plastic generation includes: The first The appropriate amount of treatment for the day and the Add up the pollution of the first day and get The amount of waste plastic generated per day.

[0016] A multivariate time series prediction system for waste plastic generation, which implements the multivariate time series prediction method for waste plastic generation, comprises: The regional setting module divides the predicted waste plastic generation area into areas covered by the sanitation system and areas not covered by the sanitation system; the plastic waste in the areas covered by the sanitation system passes through the garbage transfer station or garbage recycling facilities, and the areas not covered by the sanitation system are evenly divided into A grid; The first collection module separates the incoming garbage from the garbage transfer station into recyclable garbage and non-recyclable garbage, and collects the daily non-recyclable amount of plastic in the non-recyclable garbage; The second collection module is used to collect the recyclable amount of plastic waste from waste recycling facilities every day; The third collection module is used to collect the actual measurement of daily plastic waste in each grid in the selected grid set in the area not covered by the sanitation system; The first processing module calculates the daily amount of plastic waste in each grid in the area not covered by the sanitation system based on actual measurements; The second processing module calculates the daily disposal amount of the area covered by the sanitation system based on the non-recyclable amount and the recyclable amount; calculates the daily pollution amount of the area not covered by the sanitation system based on the calculated amount; and calculates the daily waste plastic generation amount based on the disposal amount and the pollution amount; Training modules, based on presets Set up training data and train the prediction model to predict the future The amount of waste plastic produced per day, is the length of the time window; the training data includes non-recyclable amount, recyclable amount, pollution amount, and waste plastic generation amount; The prediction module uses the time series data of non-recyclable volume, recyclable volume and pollution volume through the prediction model to predict the future The amount of waste plastic generated per day.

[0017] Beneficial effects of the multivariate time series prediction method and system for waste plastic generation of the present invention: By distinguishing between areas covered by the sanitation system and uncovered areas, a detailed statistics of plastic waste is achieved, avoiding the prediction deviation caused by factors such as the difficulty in counting the number of floating population, the difficulty in counting the amount of individual plastic waste generated, and the direct weighing of plastic waste with water and residue in the traditional method, which leads to unstable data collection deviation. The method combines the actual measurement of randomly selected grids with the minimum encirclement strategy to improve the efficiency of data collection and is more suitable for areas with frequent population mobility; the data on the plastic content in each product is directly used for calculation, which avoids the influence of water and residue on the quality of plastic waste; therefore, the present invention provides a reliable reference standard value for the prediction of the amount of waste plastic generated.

[0018] The advantage of using time series data on non-recyclable amount, recyclable amount, pollution amount and waste plastic generation to predict the amount of waste plastic generation is that non-recyclable amount, recyclable amount and pollution amount have a decisive effect on the amount of waste plastic generation. Compared with using time series data of characteristics such as population, GDP, urbanization level to predict the amount of waste plastic generation, it can significantly improve the prediction effect of the prediction model and provide strong support for plastic pollution assessment, waste disposal planning, environmental protection decision-making and public education. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a schematic diagram of the modules of the multivariate time series prediction system for waste plastic generation in Example 1 of the present invention; Figure 2 A schematic diagram of all selected grids found by the minimum encirclement strategy in Example 1 of the present invention; Figure 3 This is a schematic diagram of the minimum encirclement strategy process in Example 1 of the present invention; Figure 4 This is a schematic diagram of a module of a multivariate time series prediction system for waste plastic generation in Example 2 of the present invention; Figure 5 Schematic diagram of the flow chart of the multivariate time series prediction method for the amount of waste plastics generated according to Example 4 of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Example 1

[0022] See also Figure 1 The multivariate time series prediction system for waste plastic generation described in this embodiment includes: a regional setting module, a first acquisition module, a second acquisition module, a third acquisition module, a first processing module, a second processing module, a feature engineering module, a training module and a prediction module.

[0023] The area setting module divides the predicted waste plastic generation area into areas covered by the sanitation system and areas not covered by the sanitation system; plastic waste in the areas covered by the sanitation system passes through garbage transfer stations or garbage recycling facilities. In this embodiment, the areas covered by the sanitation system are set as areas in the preset area where regular cleaning and regular collection and transportation of garbage are arranged; the areas not covered by the sanitation system are areas in the preset area where regular cleaning or regular collection and transportation of garbage is not arranged.

[0024] The plastic waste collected and transported in the areas covered by the sanitation system are all properly handled waste, that is, they do not contain plastic waste that enters the environment and causes pollution, which is defined as the properly handled amount. The plastic waste in the areas not covered by the sanitation system are all improperly handled waste, that is, plastic waste that enters the environment and causes pollution, which is defined as the pollution amount. The data collection method and data calculation method designed by this system can respectively obtain the daily properly handled amount and daily pollution amount, and the obtained time series data of the properly handled amount and pollution amount can be further used for the training of the prediction model.

[0025] In this embodiment, it is assumed that the outbound garbage of the garbage transfer station in the area covered by the sanitation system is sent to the garbage treatment facility and the garbage recycling facility respectively; the inbound garbage of the garbage treatment facility comes only from the garbage transfer station and the garbage recycling facility; the inbound garbage of the garbage recycling facility comes from various recycling channels, including recyclable garbage sent from the garbage transfer station; the outbound garbage of the garbage recycling facility is sent to the garbage treatment facility, which is the remaining residue after the recycling process. Therefore, the proper disposal amount is equal to the mass of the plastic part of the garbage sent from the garbage transfer station to the garbage treatment facility plus the mass of the plastic part of the inbound garbage of the garbage recycling facility.

[0026] With the implementation of the EU Digital Product Passport (DPP) regulations and the smart city of Yokohama, Japan, the widespread application of RFID tags has become possible. In this embodiment, each product is provided with an RFID tag; the recorded information that can be found through the RFID tag includes the product's unique identification code, product name, production date, whether the product is recyclable, a list of parts used, and the types of plastics contained and the quality of each plastic. For example, the recorded information that can be found through the RFID tag on the mineral water bottle includes information such as "Recyclable: Yes" and "PET: 30 grams"; the recorded information that can be found through the RFID tag on the potato chip packaging bag includes information such as "Recyclable: No", "BOPP: 10 grams", "EVOH: 10 grams" and "PE: 10 grams". PET is polyethylene terephthalate, BOPP is biaxially oriented polypropylene, EVOH is ethylene-vinyl alcohol copolymer, and PE is polyethylene.

[0027] The first collection module separates the incoming garbage at the garbage transfer station into recyclable garbage and non-recyclable garbage, and collects the daily non-recyclable amount of the plastic part in the non-recyclable garbage.

[0028] The date on which this system begins to be applied is recorded as Day 1, and the day after the date on which this system begins to be applied is recorded as Day 1. The non-recyclable amount per day is recorded as The non-recyclable amount is part of the calculation of the disposal amount.

[0029] The collection methods of non-recyclable quantities include: Every day, at each garbage transfer station, the RFID tag reading device installed on the incoming garbage conveyor belt reads the RFID tag of each piece of garbage passing through; the incoming garbage is separated into recyclable garbage and non-recyclable garbage according to the recyclable information on the RFID tag; the mass of plastic garbage is determined according to the RFID tag, and the mass of the plastic part of the non-recyclable garbage after separation is calculated; the mass of the plastic part of the non-recyclable garbage in all garbage transfer stations is added up to get the mass of the plastic part of the non-recyclable garbage in the first garbage transfer station. Non-recyclable amount per day .

[0030] The second collection module is used to collect the daily recyclable plastic waste of the garbage recycling facility. The garbage recycling facility is a facility that recycles plastic waste using various recycling technologies; recycling technologies include mechanical recycling, chemical recycling, energy recycling, etc. The recyclable amount of garbage recycling facilities per day is recorded as The recyclable volume is part of the calculation of the disposal volume.

[0031] The collection methods of recyclables include: Every day, at each waste recycling facility, an RFID tag reading device installed on the inbound recyclable waste conveyor belt reads the RFID tag of each piece of waste passing through; the mass of plastic waste is determined based on the RFID tag, and the plastic portion of the inbound recyclable waste is counted; the mass of the plastic portion of the inbound recyclable waste of all waste recycling facilities is added up to obtain the mass of the plastic portion of the inbound recyclable waste of all waste recycling facilities. Recyclable volume per day .

[0032] The third collection module is used to collect the actual measurement of daily plastic waste in each grid in the selected grid set of the area not covered by the sanitation system. The grid is divided into the area not covered by the sanitation system according to the GIS information. grids (grids do not have to be square), the selected grid set is A number of grids in a grid. Each grid contains land or water.

[0033] The method for collecting the actual measurement of each grid in the selected grid set includes: Day, randomly selected The selected grid set is composed of grids; for each grid in the selected grid set, a drone carrying an RFID tag reader is used to search the land part in full along the preset path, and an unmanned submarine carrying an RFID tag reader is used to search the water part in full along the preset path; the read RFID tags are recorded, the mass of plastic waste is determined according to the RFID tags, and the mass of plastic waste in the grid is accumulated to obtain the actual measurement of each grid in the selected grid set. The preset path of the drone should cover the surface of the hill, the upper surface of the building, the ground, the road, etc. The preset path of the unmanned submarine should cover the water surface, the water, and the bottom of the water (the search can be carried out in layers by increasing the water depth by 1 meter per layer).

[0034] The first processing module calculates the daily amount of plastic waste in each grid in the area not covered by the sanitation system based on actual measurements.

[0035] The calculation method of the calculation amount includes the calculation method of the calculation amount of each grid in the selected grid set and the calculation method of the calculation amount of each grid in the unselected grid set. The day after the date on which this system begins to apply Tiandi The amount of computation in a grid is recorded as , . The amount of calculation is part of the amount of pollution.

[0036] The calculation method of the calculation amount of each grid in the selected grid set includes: Day selected grid collection The amount of calculation for each grid is equal to the actual measurement of the day, recorded as ,in The mesh number of the mesh in the selected mesh set.

[0037] like Figure 3 As shown, the calculation method for each grid in the unselected grid set will use the following minimum encirclement strategy: Step 1: Set a radius, with the geometric centroid of the currently unselected grid as the center point, and set the radius to 1 times, 2 times, 3 times, ..., Definition of radius The radius of this embodiment is set to ,in is the grid area; Step 2, start searching from an encirclement with a radius of 1. If at least one selected grid with a geometric centroid falling within the encirclement is found, record the selected grid falling within the encirclement, and then stop searching; Step 3: If no selected grid with a geometric center of mass falling within the encirclement is found, the radius of the encirclement is doubled, and step 2 is repeated until at least one selected grid with a geometric center of mass falling within the encirclement is found.

[0038] The calculation method for each grid in the unselected grid set includes: 1. At the time, The geometric centroid of the unselected grids is taken as the center point, and the average value of the actual measurements of all selected grids on the first day found by the minimum encirclement strategy is taken as the The amount of calculation for the unselected grids on the first day is recorded as ,like Figure 2 As shown, area A in the figure is Unselected grids, areas B and C are selected grids for the minimum encirclement strategy; 2. At the time, The geometric centroid of the unselected grids is taken as the center point, and the increment of the actual measurement of each selected grid on the second day relative to the first day found by the minimum encirclement strategy is calculated. The average of all increments is taken and recorded as the neighborhood increment. ; Set the first day The calculation amount of the unselected grid is added to the neighborhood increment to get the second day The computational cost of unselected grids is recorded as , ; 3. At the time, The geometric centroid of the unselected mesh is used as the center point, and the minimum enclosing circle strategy is used to calculate the The increment of the actual measurement of the day relative to the previous day is taken as the average value of the increment, which is recorded as the neighborhood increment. (The selected grid in the smallest enclosing circle is the same as the The selected grid that is closer to the first unselected grid The geographical locations and living habits of the unselected grids are relatively similar, and the average value of the increments of all selected grids in the smallest encirclement can be used to select the first Unselected grid increment calculation); Unselected grids in The calculation amount of the day is weighted averaged with the increment of the previous day and the increment of the neighborhood according to the preset weights to get the Unselected grids in The increment of the calculation amount of the day relative to the previous day is recorded as the recent increment (in this embodiment, the position of the selected grid is different for each day, considering the The recent change trend of the calculation amount of the unselected grid itself has inertia, and it is used to calculate the The accuracy of predicting recent increments can be improved when the increment of the unselected grid is Unselected grids in The calculation amount of the day is added to the recent increment to get the Unselected grids The amount of calculation per day is recorded as ; Calculate the Unselected grids The formula for calculating the amount of days is:

[0039] in, is the preset weight. In this embodiment, . Relative to For the meshes around the unselected mesh, The inertia of the recent time series data of the calculation amount of the unselected grid itself can better reflect the The variation trend of the computational amount of the unselected grid; therefore, Unselected grids in The calculation amount of the day should be given a larger weight relative to the increment of the previous day; in practical applications, experts or machine learning methods should combine the actual measurement of the grid and the difference between the calculation amount (the selected grid can get a calculation amount according to the calculation method of the unselected grid), and use the hierarchical analysis method, entropy weight method, principal component analysis method and other methods to optimize the specific value of the weight (this is the existing technology and will not be elaborated here), so that the formula of the calculation amount can be used as an estimation formula for the actual measurement.

[0040] For example: When it is 3, The computational effort of the unselected grids on the first and second days is and ;No. The number of selected grids found by using the minimum encirclement strategy for the unselected grids is 2. The actual measurements of the two selected grids on the third day are and The computational loads of the two selected grids on the second day are and , the increment of the actual measurement of the two selected grids on the third day relative to the calculated amount of the two selected grids on the second day is: , taking the average value of the increments is: , as the first The neighborhood increment of the unselected grid from the 2nd day to the 3rd day, The increase in the computational effort of the unselected grids on the second day relative to the first day is , the recent increase is , 0.8 is the preset weight, The computational effort for the third day for the unselected grid is .

[0041] The second processing module calculates the daily disposal volume in the area covered by the sanitation system based on the non-recyclable volume and the recyclable volume; calculates the daily pollution volume in the area not covered by the sanitation system based on the calculated volume; and calculates the daily waste plastic generation volume based on the disposal volume and the pollution volume. The waste plastic generation volume is the object to be predicted by the prediction module of this system.

[0042] The calculation method of the appropriate amount includes: The non-recyclable amount of the first day is Add up the recyclable amount of the day to get The appropriate amount of time for a day is recorded as , .

[0043] The calculation method of pollution amount includes: sky, , for areas not covered by the sanitation system The recent increments of the computational amount of each grid are summed to obtain The amount of pollution per day is recorded as .

[0044]

[0045] The calculation method of the amount of waste plastics generated includes: The appropriate amount of treatment for the day and the Add up the pollution of the first day and get The amount of waste plastic produced per day is recorded as , .

[0046] The feature engineering module is used to perform feature engineering operations on the time series data of non-recyclable volume, recyclable volume, pollution volume, and waste plastic production; feature engineering operations include cleaning, coding, standardization, etc. The units of non-recyclable volume, recyclable volume, pollution volume, and waste plastic production are unified as one million tons (which can be adjusted according to specific application scenarios).

[0047] Training modules, based on presets and Set up training data and train the prediction model to predict the future The amount of waste plastic produced per day, is the time window length.

[0048] Based on preset and Methods for setting up training data include: The daily characteristic data include non-recyclable amount, recyclable amount, pollution amount, and waste plastic generation amount;

[0049] The first Day to Day Heaven All feature data of the day are taken as input data, and the The daily waste plastic production is used as output data, and the input data and output data are combined into a set of training data. =3, we need to predict the amount of waste plastics generated on the next day, that is =2, the first group of training data is a group of training data consisting of input data from the 2nd to the 4th day and output data on the 6th day; the second group of training data is a group of training data consisting of input data from the 3rd to the 5th day and output data on the 7th day; the third group of training data is a group of training data consisting of input data from the 4th to the 6th day and output data on the 8th day; and so on to obtain the training data of the prediction model.

[0050] The advantage of using time series data of non-recyclable amount, recyclable amount, pollution amount and waste plastic generation to predict the amount of waste plastic generation is that the non-recyclable amount, recyclable amount and pollution amount have a decisive effect on the amount of waste plastic generation. Compared with using time series data of population, GDP, urbanization level and other characteristics to predict the amount of waste plastic generation, the prediction accuracy is significantly improved.

[0051] The training method of the prediction model includes: dividing the time series data of non-recyclable amount, recyclable amount, pollution amount and waste plastic generation amount into a training set and a test set (the ratio can be set to 7:3), using the training set to train the selected machine learning model, using the test set to test the selected machine learning model, and converging when the prediction accuracy reaches the preset prediction accuracy; finally, a prediction model that can reliably predict the amount of waste plastic generation is obtained. The machine learning model can select a long short-term memory network (LSTM) model.

[0052] The prediction module inputs the time series data of non-recyclable quantity, recyclable quantity and pollution quantity into the trained prediction model to obtain the future The amount of waste plastic generated per day.

[0053] This embodiment proposes a multivariate time series prediction system for waste plastic generation, which effectively solves the problem of inaccurate data collection on waste plastic generation in a preset area in the prior art and provides a reliable reference standard value for the waste plastic generation prediction model.

[0054] Specifically, this system achieves accurate collection of waste plastic generation data through the following methods: Data collection in areas covered by the sanitation system: Collect the daily non-recyclable amount of plastic waste in garbage transfer stations and the daily recyclable amount of plastic waste in garbage recycling facilities, and calculate the daily amount of plastic waste properly disposed of in the area covered by the sanitation system.

[0055] By combining actual measurement with calculation, the daily calculated amount of plastic waste in each grid in the area not covered by the sanitation system is obtained. Based on the calculated amount, the daily pollution amount of plastic waste in the area not covered by the sanitation system is calculated. Based on the amount of disposal and the amount of pollution, the amount of waste plastic generated in the preset area is finally calculated.

[0056] This system introduces a random measurement strategy (i.e., a strategy of randomly selecting grids for measurement) and a minimum encirclement strategy (a method of determining the averaging range), which significantly improves the ease of data acquisition while ensuring data accuracy. On the basis of providing reliable reference standard values, a machine learning model capable of predicting multivariate time series data is trained to accurately predict the amount of waste plastics generated.

[0057] Compared with the prior art, this embodiment has significant advantages: Traditional methods for collecting the amount of waste plastics are usually only applicable to areas with less population mobility, and are difficult to meet the application needs of areas with frequent population mobility. This system achieves a detailed statistics of plastic waste by distinguishing between areas covered by the sanitation system and areas not covered, avoiding the prediction bias caused by the difficulty in counting the number of mobile populations and the difficulty in accurately estimating the amount of individual plastic waste generated in traditional methods. Directly using the data on the plastic content in each product for calculation avoids the impact of water and residue on the quality of plastic waste.

[0058] Example 2

[0059] The prediction method of Example 1 cannot respond to policy changes. The present invention designs a policy intensity index as an input parameter of the prediction model. The multivariate time series prediction system for waste plastic generation in this embodiment, such as Figure 4 As shown, it also includes a fourth collection module, which is used to collect policy strength data after the policy is promulgated. The policy strength data includes strict indicators of the regional management department, plastic management budget, law enforcement intensity, etc., which can reflect the strictness of policies and regulations and the intensity of subsequent enforcement.

[0060] The third processing module calculates the policy strength index based on the policy strength data. The policy strength index is specifically calculated as follows: ; In the formula, For the The policy strength index for the day, For the The strict indicators of the day, For the Plastic management budget per day, For the The intensity of law enforcement. , It is an immediate indicator that can affect the forecast results when the policy is enacted; It is a lagging indicator that changes dynamically during the policy implementation process. In practical applications, experts or machine learning methods should be used to optimize the specific values ​​of weights based on methods such as hierarchical analysis, entropy weight method, and principal component analysis, so that the formula of the policy intensity index can be used to predict the amount of waste plastics generated under the influence of the policy. Taking the plastic restriction order as an example, the policy restricts the use of disposable plastics.

[0061] ; ; Number of penalty cases per day; In the formula, The upper limit of the fine amount is obtained through the text of the regulation; The per capita , Set aside an annual plastic management budget, The total annual public budget is obtained through public information on relevant public disclosure websites; the number of penalty cases is the number of cases recorded by functional departments that were punished for violating the plastic restriction order; the cases that were punished include a certain inspection and testing agency being punished by the market supervision and management department for issuing false inspection and testing reports, as well as the production and sale of non-degradable disposable plastic products; The value-taking methods include: Re-value before predictive model training to track the latest data changes; after the predictive model is trained until the next predictive model training, The value of remains unchanged to maintain the uniformity of data normalization and the effectiveness of prediction; when the prediction model is retrained, The maximum value of the historical data is twice as high as Similarly, we can get and The value-taking method of .

[0062] By capping the amount of fines ( ) to GDP per capita to reflect the strict indicators and severity of plastic management policies. The ratio of the annual plastic management budget (SG) to the annual total public budget (ZG) reflects the degree of policy funding bias of local management departments in implementing plastic management, and whether there are sufficient resources to support the implementation of plastic restriction policies; sufficient resource support will improve the efficiency of policy implementation. The intensity of law enforcement is reflected through the number of cases punished daily. The number of cases punished daily is a direct indicator to measure the frequency and effectiveness of law enforcement activities. It can reflect the intensity of policy implementation. The higher the intensity of implementation, the more it helps to ensure that the policy is effectively implemented and reduce the occurrence of violations.

[0063] The daily characteristic data also includes a policy strength index.

[0064] This embodiment takes into account that policy changes may cause short-term impacts. By introducing a policy intensity index, it is possible to respond quickly to policy changes and improve the prediction accuracy of the prediction model in the event of policy changes.

[0065] Example 3

[0066] The multivariate time series prediction system for waste plastic generation in this embodiment provides another method for collecting actual measurements, as follows: The drone is used to conduct a complete search along the preset path to obtain image data on the preset path, and the obtained image data is input into the trained plastic recognition model to obtain the type of plastic. The mass of the plastic is determined based on the type of plastic, and the mass of the plastic waste in the grid is accumulated to obtain the actual measurement of each grid in the selected grid set.

[0067] The training method of the plastic recognition model includes: Collect image data and the corresponding plastic type annotations in the image data in advance, use the image data as the input of the plastic recognition model, and the plastic recognition model uses the plastic type annotations actually corresponding to the image data as output. Use loss functions and optimizers to train the plastic recognition model, such as cross entropy loss function and Adam optimizer. During the training process, use data enhancement technology to increase data diversity and improve the generalization ability of the model, and complete the training until the preset prediction accuracy of the plastic recognition model is achieved. The plastic recognition model is specifically a convolutional neural network (CNN) or a YOLOv8 model.

[0068] Methods for determining the mass of plastic waste by type of plastic include: Input the type of plastic into the quality query database to obtain the mass of the plastic corresponding to the plastic type. For example, if the plastic type is a certain brand of mineral water bottle or plastic packaging bag, the quality query database will record the mass corresponding to each type of plastic in advance. You can query the specific name of the plastic type in the quality query database.

[0069] Example 4

[0070] like Figure 5As shown, this embodiment provides a multivariate time series prediction method for the amount of waste plastic generated, including: dividing the predicted waste plastic generation area into an area covered by the sanitation system and an area not covered by the sanitation system; the plastic waste in the area covered by the sanitation system passes through a garbage transfer station or a garbage recycling facility, and the area not covered by the sanitation system is evenly divided into A grid; Separate the incoming garbage at the garbage transfer station into recyclable and non-recyclable garbage, and collect the daily non-recyclable amount of plastic in the non-recyclable garbage; Collect the recyclable amount of plastic waste from recycling facilities every day; Collect daily measurements of plastic waste for each grid in a selected set of grids in areas not covered by sanitation systems; Calculate the amount of daily plastic waste in each grid in areas not covered by the sanitation system based on actual measurements; Calculate the daily disposal volume in areas covered by the sanitation system based on the non-recyclable volume and the recyclable volume; calculate the daily pollution volume in areas not covered by the sanitation system based on the calculated volume; calculate the daily waste plastic generation volume based on the disposal volume and the pollution volume; Based on preset and Set up training data and train the prediction model to predict the future The amount of waste plastic produced per day, is the length of the time window; the training data includes non-recyclable amount, recyclable amount, pollution amount, and waste plastic generation amount; Using the time series data of non-recyclable, recyclable and polluted quantities through the prediction model, the future The amount of waste plastic generated per day.

[0071] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0072] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A multivariate time series prediction method for waste plastic generation, characterized in that: include: The predicted waste plastic generation areas are divided into areas covered by the sanitation system and areas not covered by the sanitation system; Plastic waste in areas covered by the sanitation system passes through waste transfer stations or waste recycling facilities. Areas not covered by the sanitation system are divided equally into A grid; Separate the incoming garbage at the garbage transfer station into recyclable and non-recyclable garbage, and collect the daily non-recyclable amount of plastic in the non-recyclable garbage; Collect the recyclable amount of plastic waste from recycling facilities every day; Collect daily measurements of plastic waste for each grid in a selected set of grids in areas not covered by sanitation systems; Calculate the amount of daily plastic waste in each grid in areas not covered by the sanitation system based on actual measurements; Calculate the daily disposal volume in areas covered by the sanitation system based on the non-recyclable volume and the recyclable volume; calculate the daily pollution volume in areas not covered by the sanitation system based on the calculated volume; calculate the daily waste plastic generation volume based on the disposal volume and the pollution volume; Based on preset Set up training data and train the prediction model to predict the future The amount of waste plastic produced per day, is the time window length; The training data includes non-recyclable amount, recyclable amount, pollution amount, and waste plastic generation amount; Using the time series data of non-recyclable, recyclable and polluted quantities through the prediction model, the future The amount of waste plastic generated per day.

2. The multivariate time series prediction method for waste plastic generation according to claim 1, characterized in that: The calculation amount includes the calculation amount of each grid in the selected grid set and the calculation amount of each grid in the unselected grid set. Day selected grid collection The computational effort for each grid is equal to the actual measurement of the day; the computational effort for each grid in the unselected grid set is calculated based on the actual measurement of the selected grid found by the minimum encirclement strategy.

3. The multivariate time series prediction method for waste plastic generation according to claim 2, characterized in that: The methods for calculating the amount of each mesh in the unselected mesh set include: At the time, The geometric centroid of the unselected mesh is the center point. The average value of the actual measurements of all selected grids on the first day found using the minimum encirclement strategy is taken as the The amount of calculation for the unselected grids on the first day; At the time, The geometric centroid of the unselected grids is taken as the center point, and the increment of the actual measurement of each selected grid on the second day relative to the first day found by the minimum encirclement strategy is calculated. The average of all increments is taken and recorded as the neighborhood increment. The calculation amount of the unselected grid is added to the neighborhood increment to get the second day The amount of calculation for unselected grids; At the time, The geometric centroid of the unselected mesh is used as the center point, and the minimum enclosing circle strategy is used to calculate the The increment of the actual measurement of the day relative to the previous day is taken, and the average value of the increment is recorded as the neighborhood increment; Unselected grids in The calculation amount of the day is weighted averaged with the increment of the previous day and the increment of the neighborhood according to the preset weights to get the Unselected grids in The increment of the calculation amount of the day relative to the previous day is recorded as the recent increment; Unselected grids in The calculation amount of the day is added to the recent increment to get the Unselected grids Days of calculations.

4. The multivariate time series prediction method for waste plastic generation according to claim 3, characterized in that: Minimum encirclement strategies include: Step 1: Set a radius, take the geometric centroid of the currently unselected grid as the center point, and gradually increase the radius by multiples to form a circular encirclement; Step 2, start searching from the encirclement with the minimum multiple radius. If at least one selected grid with a geometric centroid falling within the encirclement is found, record the selected grid falling within the encirclement, and then stop searching; Step 3: If no selected grid with a geometric center of mass falling within the encirclement is found, the radius of the encirclement is increased by 1 times based on the current radius, and step 2 is repeated until at least one selected grid with a geometric center of mass falling within the encirclement is found.

5. The multivariate time series prediction method for waste plastic generation according to claim 1, characterized in that: Based on preset and Methods for setting up training data include: The first Day to Day All feature data of the day are taken as input data, and the The daily waste plastic generation is used as the output data, and the input data and output data are combined into a set of training data. The daily feature data includes the non-recyclable amount, recyclable amount, pollution amount, and waste plastic generation.

6. The multivariate time series prediction method for waste plastic generation according to claim 1, characterized in that: The area covered by the sanitation system is the area in the preset area where regular cleaning and regular collection and transportation of garbage are arranged; the area not covered by the sanitation system is the area in the preset area where regular cleaning or regular collection and transportation of garbage are not arranged; The outbound garbage from the garbage transfer stations in the area covered by the sanitation system is sent to two directions: garbage treatment facilities and garbage recycling facilities; the inbound garbage to the garbage treatment facilities comes only from garbage transfer stations and garbage recycling facilities; the inbound garbage to the garbage recycling facilities includes recyclable garbage sent from the garbage transfer stations; the outbound garbage from the garbage recycling facilities is sent to the garbage treatment facilities and is the remaining residue after the recycling process.

7. The multivariate time series prediction method for waste plastic generation according to claim 1, characterized in that: The method for collecting the non-recyclable amount includes: In the Every day, at each garbage transfer station, the RFID tag reading device installed on the incoming garbage conveyor belt reads the RFID tag of each piece of garbage passing through; the incoming garbage is separated into recyclable garbage and non-recyclable garbage according to the recyclable information on the RFID tag; the mass of plastic garbage is determined according to the RFID tag, and the mass of the plastic part of the non-recyclable garbage after separation is calculated; the mass of the plastic part of the non-recyclable garbage in all garbage transfer stations is added up to get the mass of the plastic part of the non-recyclable garbage in the first garbage transfer station. The non-recyclable amount of the day.

8. The multivariate time series prediction method for waste plastic generation according to claim 1, characterized in that: The method for collecting the recyclable amount includes: In the Every day, at each waste recycling facility, an RFID tag reading device installed on the inbound recyclable waste conveyor belt reads the RFID tag of each piece of waste passing through; the mass of plastic waste is determined based on the RFID tag, and the plastic portion of the inbound recyclable waste is counted; the mass of the plastic portion of the inbound recyclable waste of all waste recycling facilities is added up to obtain the mass of the plastic portion of the inbound recyclable waste of all waste recycling facilities. Recyclable volume per day; The method for collecting the actual measurement of each grid in the selected grid set includes: In the Day, randomly selected The selected mesh collection consists of meshes. ; For each grid in the selected grid set, a drone carrying an RFID tag reader is used to conduct a complete search along a preset path in the land part, and an unmanned submarine carrying an RFID tag reader is used to conduct a complete search along a preset path in the water part; the read RFID tags are recorded, the mass of the plastic waste is determined based on the RFID tags, the mass of the plastic waste in the grid is accumulated, and the actual measurement of each grid in the selected grid set is obtained.

9. The multivariate time series prediction method for waste plastic generation according to claim 1, characterized in that: The calculation method of the appropriate amount includes: The first The non-recyclable amount of the first day is Add up the recyclable amount of the day to get The proper measure of heaven; The method for calculating the pollution amount includes: In the sky, , for areas not covered by the sanitation system The recent increments of the computational amount of each grid are summed to obtain The amount of pollution per day; The calculation method of waste plastic generation includes: The first The appropriate amount of treatment for the day and the Add up the pollution of the first day and get The amount of waste plastic generated per day.

10. A multivariate time series prediction system for waste plastic generation, characterized in that: For implementing the multivariate time series prediction method for waste plastic generation according to any one of claims 1 to 9, the system comprises: The regional setting module divides the predicted waste plastic generation area into areas covered by the sanitation system and areas not covered by the sanitation system; the plastic waste in the areas covered by the sanitation system passes through the garbage transfer station or garbage recycling facilities, and the areas not covered by the sanitation system are evenly divided into A grid; The first collection module separates the incoming garbage from the garbage transfer station into recyclable garbage and non-recyclable garbage, and collects the daily non-recyclable amount of plastic in the non-recyclable garbage; The second collection module is used to collect the recyclable amount of plastic waste from waste recycling facilities every day; The third collection module is used to collect the actual measurement of daily plastic waste in each grid in the selected grid set in the area not covered by the sanitation system; The first processing module calculates the daily amount of plastic waste in each grid in the area not covered by the sanitation system based on actual measurements; The second processing module calculates the daily disposal amount of the area covered by the sanitation system based on the non-recyclable amount and the recyclable amount; calculates the daily pollution amount of the area not covered by the sanitation system based on the calculated amount; and calculates the daily waste plastic generation amount based on the disposal amount and the pollution amount; Training modules, based on presets Set up training data and train the prediction model to predict the future The amount of waste plastic produced per day, is the length of the time window; the training data includes non-recyclable amount, recyclable amount, pollution amount, and waste plastic generation amount; The prediction module uses the time series data of non-recyclable volume, recyclable volume and pollution volume through the prediction model to predict the future The amount of waste plastic generated per day.

Images