Urban kitchen waste production and discharge prediction method based on BPN-RF

Abstract

The invention relates to the technical field of kitchen waste treatment, in particular to an urban kitchen waste production and emission prediction method based on BPN-RF, which comprises the following steps: acquiring input data of seat number, service number, raw materials and turnover and output data corresponding to kitchen waste; the BPN network is trained; performing kitchen waste value prediction on the trained BPN; performing initial interval division, and performing secondary interval division according to the initial interval division; taking the secondary interval division and the corresponding seat number, per capita kitchen waste, per capita catering consumption and local per capita catering consumption as input data, taking secondary interval division data as output data, and training an RF algorithm; and predicting the trained RF algorithm by using the seat number, the per capita kitchen waste, the per capita catering consumption and the local per capita catering consumption to obtain a kitchen waste interval prediction value. According to the method, the problems of poor robustness and insufficient precision of the existing K-means-RBF method are solved.

Description

Technical Field

[0001] The invention relates to the technical field of restaurant kitchen waste treatment, and in particular to a method for predicting the generation and discharge of urban restaurant kitchen waste based on BPN-RF. Background Art

[0002] The amount of food waste generated increases year by year, bringing a series of problems to the city; Shushe et al. further predicted the spatial distribution of food waste generated at the street scale in Xiamen based on regression analysis; the ARIMA model is the most commonly used time series analysis method, which has good accuracy in short-term predictions, but lacks interpretability in internal connections; the grey system is applied to the prediction of waste generation; its prediction has been proven to have good performance in the absence of data, but it can only be applied to common scenarios under a few limited conditions; existing research on solid waste generation issues is mostly focused on predicting waste generation based on official statistical bulletins, lacking actual investigation and combination with spatial distribution based on geographic information system (GIS). Although there is a qualitative analysis of the internal influencing factors of the generation characteristics, there is a lack of effective classification and quantitative evaluation of key factors.

[0003] The patent with publication number CN 113379569A adopts a combination of K-means clustering and RBF neural network to perform interval prediction of restaurant kitchen waste. First, RBF and K-means clustering methods have poor robustness and insufficient precision. Secondly, this method divides static data into intervals and then uses one-hot encoding. The model outputs discrete production and discharge intervals. The discrete intervals have insufficient precision, that is, the discrete interval division will cause errors. It is also not intuitive enough. Finally, this method is applied to specific areas, and garbage collection and transportation are guided by heat maps and grid analysis methods, which has the problem of insufficient flexibility. Summary of the invention

[0004] In view of the shortcomings of the existing methods, the present invention solves the problems of poor robustness and insufficient precision of the existing K-means-RBF method.

[0005] The technical solution adopted by the present invention is: the method for predicting the generation and discharge of urban kitchen waste based on BPN-RF comprises the following steps:

[0006] Step 1: Collect input data of number of seats, number of services, raw materials, turnover and output data of corresponding restaurant kitchen waste;

[0007] Step 2: Use the data from step 1 to train the BPN network; use the assumed number of seats, number of services, raw materials, and turnover to predict the value of restaurant kitchen waste using the trained BPN network;

[0008] As a preferred embodiment of the present invention, the assumed number of seats, number of services, raw materials and turnover are expanded by combining Cartesian products.

[0009] Step 3: Perform initial interval division according to the frequency of the predicted kitchen waste value, and perform secondary interval division according to the initial interval division;

[0010] As a preferred embodiment of the present invention, the initial interval division is performed based on the occurrence frequency of the kitchen waste value calculated in the first interval.

[0011] As a preferred implementation manner of the present invention, the initial interval division includes a first frequency zone, a second frequency zone and a third frequency zone.

[0012] As a preferred implementation of the present invention, the frequency of the first frequency zone is greater than or equal to the first frequency threshold; the frequency of the third frequency zone is less than the third frequency threshold; and the frequency of the second frequency zone is between the first frequency threshold and the third frequency threshold.

[0013] Step 4: Using the secondary interval division and the corresponding number of seats, per capita restaurant kitchen waste, per capita catering consumption, and local per capita catering consumption as input data and the secondary interval division data as output data, the RF algorithm is trained;

[0014] As a preferred embodiment of the present invention, the secondary interval division is to set a second interval when the cumulative distribution value of the food waste after the initial interval division is less than the first food waste threshold.

[0015] As a preferred embodiment of the present invention, the secondary interval division is to set the fourth interval when the cumulative distribution value of the food waste after the initial interval division is greater than the second food waste threshold.

[0016] As a preferred embodiment of the present invention, the secondary interval division is to set the third interval when the cumulative distribution value of the food waste after the initial interval division is greater than or equal to the first food waste threshold and less than or equal to the second food waste threshold.

[0017] Step 5: Use the number of seats, per capita restaurant kitchen waste, per capita restaurant consumption, and local per capita restaurant consumption to predict the trained RF algorithm and obtain the predicted value of the restaurant kitchen waste interval;

[0018] As a preferred embodiment of the present invention, a system for predicting the generation and discharge of urban restaurant kitchen waste based on BPN-RF includes: a memory for storing instructions executable by a processor; and a processor for executing instructions to implement a method for predicting the generation and discharge of urban restaurant kitchen waste based on BPN-RF.

[0019] As a preferred embodiment of the present invention, a computer readable medium stores a computer program code, and when the computer program code is executed by a processor, a method for predicting the generation and discharge of urban food waste based on BPN-RF is implemented.

[0020] Beneficial effects of the present invention:

[0021] 1. The machine learning model that combines BPN and RF is used. The BPN can process nonlinear data and the RF can process high-dimensional data and avoid overfitting. The model combination has higher robustness and accuracy.

[0022] 2. Use continuous features for interval division. The machine learning model outputs continuous values, which is more accurate and intuitive, avoiding the errors that may be caused by interval division.

[0023] 3. The present invention is applicable to the management of urban kitchen waste of different sizes and types, and has strong deployment versatility through flexible GIS presentation. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a prediction logic diagram of urban kitchen waste generation and discharge based on BPN-RF of the present invention;

[0025] Figure 2 It is a BPN network structure diagram of the present invention;

[0026] Figure 3 is the RF structure diagram of the present invention;

[0027] Figure 4 is the linear fitting and correlation heat map of the present invention;

[0028] Figure 5 It is the frequency and kitchen waste distribution diagram of the present invention;

[0029] Figure 6 It is a simulated change trend diagram of the dynamic data and the amount of kitchen waste generated in the present invention;

[0030] Figure 7 It is the distribution diagram of the kitchen waste generation of the present invention;

[0031] Figure 8 It is the characteristic curve of kitchen waste generation of different types of catering units of the present invention;

[0032] Fig. 9 It is the material flow diagram of the Chinese restaurant of the present invention. DETAILED DESCRIPTION

[0033] The present invention is further described below in conjunction with the accompanying drawings and embodiments. This figure is a simplified schematic diagram, which only illustrates the basic structure of the present invention in a schematic manner, and therefore it only shows the components related to the present invention.

[0034] like Figure 1 As shown in FIG. 1 , the method for predicting the generation and discharge of urban kitchen waste based on BPN-RF includes the following steps:

[0035] Step 1: Collect input data of number of seats, number of services, raw materials, turnover and output data of corresponding restaurant kitchen waste;

[0036] The input data include quantitative indicators in the fields of economy, urbanization, life, etc. The quantitative indicators can be directly measured or obtained from official document statistics, announcements, etc.; while social demographic factors, personal factors, natural factors and policy factors are qualitative indicators in most cases, and these indicators need to be quantified according to given circumstances; in contrast, intrinsic factors can most directly reflect the situation of garbage generation; the present invention takes intrinsic factors as a premise, considers the spontaneous corrections brought about by social demographic factors and personal factors, and the mandatory corrections brought about by natural factors and policy factors to predict the amount of restaurant kitchen waste generated; the use of linear regression models and correlation analysis can reflect the impact of survey data on the amount of restaurant kitchen waste generated, which is divided into static data and dynamic data.

[0037] Among them, the number of seats, per capita food waste generation, per capita catering consumption, per capita catering consumption, and local per capita catering consumption are static data; the number of services, raw materials, and turnover are dynamic data;

[0038] The number of services refers to the number of services provided by a catering service unit during the business hours of a day; raw materials refer to the amount of various ingredients and materials used by a restaurant to prepare dishes in a day; and turnover refers to all the income a restaurant receives from customers purchasing catering services and goods in a day.

[0039] The number of seats, services, raw materials, turnover, and restaurant kitchen waste in step 1 are survey data;

[0040] The survey data is obtained by collecting objective and actual information through questionnaire surveys; static data reflects the business type of the restaurant and can be used to approximately predict the operating conditions of different restaurants; dynamic data is used to simulate the performance of restaurants in different seasons; in this embodiment, the survey data time is from March to December 2023, and a total of 693 businesses of various types are randomly surveyed in commercial streets and shopping malls in various districts of a certain state.

[0041] Step 2: Use the data from step 1 to train the BPN network; use the assumed number of seats, number of services, raw materials, and turnover to predict the value of restaurant kitchen waste using the trained BPN network;

[0042] The hypothetical data is an expansion of the survey data; as shown in Table 1, the hypothetical data for 693 restaurants based on the survey data;

[0043] Table 1 The selected interval values ​​and investigation results of the number of seats and dynamic data

[0044]

[0045]

[0046] As shown in Table 1, there are 11 assumed classifications for the number of seats, and 17 assumed classifications for the number of services, turnover, and raw materials. Combining the four assumed data using Cartesian product yields a total of 12*17*17*17=58956 combinations of assumed data.

[0047] like Figure 2 The BPN network structure diagram is shown in Figure 2. BPN is a multi-layer feedforward neural network whose weights are randomly set initially and can be continuously modified during back propagation until the difference between the training data and the output data meets the final value of the training accuracy. The formula for the hidden layer is:

[0048] y=b+Σx i w i

[0049] Where b is the bias, x is i is the initial variable value, w i is the weight.

[0050] Using the decisive factor (R 2 ) and mean absolute percentage error (MAPE) to evaluate the accuracy of the BPN network, the formula is:

[0051]

[0052] Among them, y i is the actual value, is the predicted value.

[0053] Overall Accuracy (OA) is the proportion of correctly predicted samples to all samples, and the formula is:

[0054]

[0055] Among them, TP is true positive, TN is true negative, FN is false negative, and FP is false positive.

[0056] The formula for the linear regression model is:

[0057] y i =a+bx i

[0058] In the formula, x i With y i are the independent variable and the dependent variable, a is the constant term, and b is the regression coefficient;

[0059] The formula for correlation analysis is:

[0060]

[0061] In the formula, and is the sample mean, r is the correlation coefficient, and n is the sample size.

[0062] The use of linear regression model and correlation analysis can reflect the impact of survey data on the amount of food waste generated, and the structure and parameters of the neural network can be adjusted through the results of linear regression and correlation analysis; the selection of the input layer of the BPN network is based on the results of correlation analysis, and input features that are highly correlated with the output variables are selected to reduce the complexity of network training; weight initialization uses the coefficients of linear regression as the initial values ​​of the neural network weights, because there is already a reasonable basis for its initialization, which can accelerate the convergence process of the network; adjustment of learning rate and number of iterations: linear regression provides a good initial fit, and the learning rate and maximum number of iterations of the neural network are adjusted based on this result; the fitting error of the linear regression model can be used as a reference for network optimization, and a reasonable learning rate can be set to prevent convergence that is too fast or too slow; model selection: adjust the activation function of the neural network according to the actual effect of the model, for example: for tasks with strong linear relationships, you can try to use linear activation functions; for tasks with strong nonlinear relationships, you can try nonlinear activation functions such as Sigmoid and ReLU.

[0063] Figure 4 a is a linear fitting diagram of the survey data and the amount of restaurant kitchen waste generated. The results show that raw materials have the greatest impact on the amount of restaurant kitchen waste generated (R 2 =0.82), followed by turnover, number of seats and number of services, and the impact of per capita catering consumption and per capita catering kitchen waste on the amount of catering kitchen waste generation was not significant; Figure 4 b is a heat map of the correlation between survey data and the amount of restaurant kitchen waste generated. The results show that raw materials, turnover, number of seats, number of services, per capita catering consumption, per capita restaurant kitchen waste and the amount of restaurant kitchen waste generated are positively correlated, while the local per capita catering consumption is negatively correlated with the amount of restaurant kitchen waste generated, which is consistent with the actual situation; restaurants that generate more restaurant kitchen waste serve more guests and have higher requirements for the supply capacity of raw materials, while per capita catering consumption and per capita restaurant kitchen waste reflect the grade of the restaurant and the consumption habits of the customers, and have a weaker impact than other factors.

[0064] Since dynamic data has a significant impact on the amount of restaurant kitchen waste generated, it is further analyzed; Figure 6 Display dynamic data and simulated change trend chart of food waste generation; Figure 6 a is the number of services, turnover, and raw materials; Figure 6 The b is the number of seats, raw materials, and turnover; Figure 6 The c is the number of seats, services, and turnover; Figure 6d is the number of seats, the number of services, and the raw materials; the color and size of the data points in the bubble chart reflect the amount of restaurant kitchen waste generated. It can be seen that under the same conditions, restaurant kitchen waste increases significantly with the increase of raw materials, and restaurant kitchen waste is positively correlated with the supply capacity of raw materials. This is similar to the analysis results of the survey data. In the actual survey, high-end and medium-to-large restaurants generate more food waste. Their dishes are rich in variety, the amount of ingredients used is large, and consumers are accustomed to ordering more dishes at one time, which easily causes more waste; when other parameters remain unchanged, restaurant kitchen waste increases with the increase in the number of services, but does not change significantly with the number of seats and turnover, which is slightly different from the changing trend of the survey data. This may be because the operating conditions of some restaurants fluctuate in real life, resulting in the change of the number of seats not significantly affecting the amount of restaurant kitchen waste generated. It may also be because the take-out behavior of some restaurants leads to a decrease in restaurant kitchen waste. Although the turnover has increased, the amount of restaurant kitchen waste generated has not increased; raw materials and the number of services have a greater impact on the amount of restaurant kitchen waste generated.

[0065] Step 3: Perform initial interval division according to the frequency of the predicted kitchen waste value, and perform secondary interval division according to the initial interval division;

[0066] In this embodiment, the R 2 and MAPE are 0.86 and 5.58 respectively, indicating that the model has high accuracy;

[0067] The initial interval division is based on the frequency of occurrence of the kitchen waste value calculated in the first interval; the first interval is set to 0.1 kg.

[0068] The initial interval division includes the first frequency zone (high frequency zone), the second frequency zone (middle frequency zone) and the third frequency zone (low frequency zone);

[0069] The frequency of the first frequency region is greater than or equal to the first frequency threshold; the frequency of the third frequency region is less than the third frequency threshold; and the frequency of the second frequency region is between the first frequency threshold and the third frequency threshold.

[0070] In this embodiment, the first frequency threshold = 180, the second frequency threshold is between the first frequency threshold and the third frequency threshold, which is 5-180; the third frequency threshold <= 5;

[0071] Figure 5 The distribution of a is divided according to the first interval and three frequency thresholds; the first interval is the frequency distribution of 0.1kg; the peak values ​​of the high-frequency and medium-frequency areas fluctuate between 12.8kg and 44kg, and there is no peak in the low-frequency area; the peak value of 17.7 in the high-frequency area is rounded off to obtain 17.7kg and the corresponding frequency is 294.

[0072] Figure 5Figure b shows the frequency distribution when the food waste value interval is divided into 15kg intervals; the catering units with a daily food waste generation of 15-30kg accounted for the largest proportion (44.78%), followed by 0-15kg (39.88%), while the catering units with a daily food waste generation of 30-240kg accounted for only 15.29%; Figure 5 c shows the frequency distribution of the asymptotic interval of restaurant kitchen waste values. Catering units with a daily kitchen waste generation of 0-45kg account for 93.45%, while those with a daily kitchen waste generation of 240-1065kg account for only about 0.2%, which is consistent with the actual survey situation; most of the restaurants with a daily kitchen waste generation of 0-45kg are small restaurants such as fast food restaurants and snack bars, which require raw material supply capabilities that are close to their raw material consumption capabilities, have a large number, and are relatively concentrated in distribution. The total amount of kitchen waste generated is large, and they are an important component of the city's kitchen waste generation; while restaurants with a daily kitchen waste generation of more than 45kg are mainly medium and large restaurants, which have a large demand for raw materials, a relatively small number, and a dispersed overall distribution; the simulation results more objectively reflect the numerical distribution trend of the city's kitchen waste generation, and can provide a reference for the prediction and scientific monitoring and management of the city's kitchen waste generation.

[0073] Step 4: Using the secondary interval division and the corresponding number of seats, per capita restaurant kitchen waste, per capita catering consumption, and local per capita catering consumption as input data and the secondary interval division data as output data, the RF algorithm is trained;

[0074] like Figure 3 It is the RF algorithm. RF is an integrated learning method that uses the bagging algorithm and random features to build decision trees and output their average predictions.

[0075] The secondary interval division is to set the second interval when the cumulative distribution value of the food waste after the initial interval division is less than the first food waste threshold;

[0076] It also includes: when the cumulative distribution value of the kitchen waste is greater than the second kitchen waste threshold, setting a fourth interval;

[0077] It also includes: when the first kitchen waste threshold = < the cumulative distribution value of kitchen waste < = the second kitchen waste threshold, setting the third interval.

[0078] from Figure 5From b and c, we can see that the 0-45kg interval is a high frequency, and the corresponding maximum proportion of the cumulative distribution value is 95%, that is, the first kitchen threshold = 95%: this interval contains the garbage volume of most catering units; the interval can be refined into smaller sub-intervals, and the second interval is set to 5kg; in order to capture more subtle changes; this can improve the accuracy of the prediction and reduce the model error, especially in the prediction of low garbage volume range. The corresponding intervals are divided into: 0-5kg, 6-10kg, 11-15kg, ..., 41-45kg. After the subdivision interval, the changes in the garbage volume in a small range can be captured, providing more accurate support for garbage prediction in this interval;

[0079] 45kg-240kg is a low frequency, and the corresponding maximum cumulative distribution value is 98%, that is, the second kitchen threshold is 98%, and the corresponding third interval between 95% and 98% is set to 50kg;

[0080] The 240-1065kg interval is an extremely low frequency, that is, it is greater than the second kitchen threshold of 98%, indicating that only a few restaurants have garbage falling into this range. In this case, too detailed a division of this interval may cause the model to be too complex and increase unnecessary computational costs. Therefore, appropriately relaxing the interval division of this part can reduce the error caused by excessive division and avoid overfitting of a small number of samples. The fourth interval is set to 500kg, and the interval divisions are such as 240-740kg and 741-1065kg. After relaxing the interval, although the data in each interval may not be as dense as when it is finely divided, this will avoid overfitting of the model due to too many divisions while maintaining a relatively reasonable prediction accuracy.

[0081] The results of BPN simulation prediction can provide assistance for reasonable interval division and effectively balance the division accuracy of low-frequency intervals and high-frequency intervals, so that the model can maintain good generalization ability and prediction accuracy in the entire garbage prediction range.

[0082] Step 5: Use the number of seats, per capita restaurant kitchen waste, per capita restaurant consumption, and local per capita restaurant consumption to predict the trained RF algorithm and obtain the predicted value of the restaurant kitchen waste interval;

[0083] The official data and POI data were used to obtain the number of seats in each restaurant, per capita food waste, per capita catering consumption, and local per capita catering consumption.

[0084] Official data refers to data publicly released by statistical departments and other organizations; POI data is obtained by crawling data from third-party consumer review websites using Python.

[0085] Taking into account the influence of the variable area unit problem (MAUP), a spatial grid was constructed as the analysis unit. The Fishnet tool of ArcGIS software was used to construct a square grid with a side length of 1 km covering the urban area of ​​the city. The grid was used as a statistical unit to calculate the density of the catering industry and other indicators through spatial association and other methods for data processing and variable aggregation. Each restaurant was located through POI and geographic information, and an urban grid with a resolution of 1km×1km was constructed. Its spatial distribution was visualized using Arc-GIS software. The material flow of the restaurant was drawn using Sankey software.

[0086] Figure 7 a is the spatial distribution of restaurant kitchen waste in the city with a 1km×1km area drawn based on a fishing net; the overall accuracy (OA) of the RF model reaches 0.85; by summarizing the POI prediction data in each grid, it is obtained that the daily restaurant kitchen waste generated by 53,300 restaurants in the city is 1143.53t; from the perspective of the amount generated in each district, the regional restaurant kitchen waste accounts for 12.5%, 27.90%, 19.19%, 13.73%, 13.61% and 13.05% in Jintan District, Wujin District, Xinbei District, Tianning District, Zhonglou District and Liyang City respectively; Wujin District, Tianning District and Zhonglou District are connected and have a high concentration of restaurants, forming a restaurant kitchen waste distribution area with a high concentration of restaurants. In the high-density kitchen waste area, Jintan District and Liyang City have also formed a high-density area in the central area, while Xinbei District presents a discrete medium-density area; in this regard, the high-density area can take advantage of the dense kitchen waste to promote garbage classification, unified collection and transportation, and centralized resource disposal, and send more and larger-capacity collection and transportation vehicles to reduce transportation costs; implement multi-department joint supervision and management, with clear division of labor and responsibility, and establish a three-level management network in which the Municipal Environmental Sanitation Bureau, District Environmental Sanitation Bureau, and Street Environmental Sanitation Office each take responsibility, and realize efficient management of kitchen waste in the whole society through source classification management in the generation link, standardized management in the collection and transportation link, and market operation in the disposal link; Figure 7b compares the generation of restaurant kitchen waste in each district and the influencing factors including the population, GDP, number of restaurants, catering industry income and per capita catering consumption of each district; Wujin District and Xinbei District account for nearly half (47.09%) of the city's restaurant kitchen waste generation. The main reason for this phenomenon is that they have a large population and catering units, and their GDP is also at the forefront; from the figure, it can be found that the proportion of catering units in the central urban area (Xinbei District, Tianning District, Zhonglou District) is higher than the proportion of population, while the proportion of catering units in Wujin District, Jintan District and Liyang City is lower than the proportion of population. The main reason for this phenomenon is that the central urban area is the city's administrative, economic, and foreign cultural exchange center, and the business district is developed; the main functions of other regions are education, industry, and tourism, and the proportion of the service industry is relatively low. Therefore, the catering industry is relatively small in the north, and the distribution area is limited and concentrated, mainly around residential areas and business districts; these are consistent with the results of Abiodun's study, population size, GDP and other factors related to urbanization The changes will promote the increase in the number of restaurants and the frequency of dining out; catering income reflects the local catering business conditions, and per capita catering consumption represents the local personal consumption potential. These two factors are most prominent in Tianning District, but its food waste generation ranks only third. It can be found that factors such as catering income and per capita catering consumption have no significant impact on the amount of urban food waste generation, which is consistent with the analysis results of the survey data; high-resolution food waste spatial map can provide a reference for the site selection, service scale and capacity of treatment plants in decentralized food waste treatment systems, and is also conducive to the optimization of collection and transportation routes; according to field research, it is learned that the two food waste treatment plants in the city are about to reach their load limit. In the future, the planning and layout of food waste collection and transportation points and related infrastructure can be built around the central urban area to reduce collection and transportation costs. Other discrete districts can reduce costs and increase efficiency by installing decentralized food waste treatment facilities or by jointly treating food waste with other organic waste.

[0087] Based on the survey data, POI data and RF prediction results, the relationship between the generation of food waste from different types of catering units in the city and the number of restaurants and per capita emissions was drawn, such as Figure 8; According to the POI data classification, catering units are divided into 9 categories, namely Chinese food, fast food, Western food, buffet, bread and dessert, dessert and beverage, hot pot, southeastern cuisine, Japanese and Korean cuisine; fast food restaurants and Chinese restaurants account for 42.5% and 37.8% of the total number of catering units in the city respectively, followed by dessert and beverage shops with 8.04% and hot pot restaurants with 5.17%; Chinese restaurants, fast food restaurants and hot pot restaurants are the three types of restaurants that generate the most daily food waste; among them, fast food, dessert and beverage, bread and dessert, etc., due to the characteristics of simple dishes and small amount of ingredients, not only the amount of food waste generated is not proportional to the number of catering stores, but also the per capita generation is low; Chinese restaurants generate the most food waste, these restaurants offer a wide variety of dishes and rich ingredients, and the amount of food waste is large; although there are not many hot pot restaurants, they will produce a large amount of high-oil mixture when dining, and the per capita generation is the largest; dessert and beverage shops generate the most food waste It mainly comes from the remaining liquid. In addition, beverage shops mainly undertake take-out services and staff sometimes pour liquid waste directly into the drain, so the amount generated may fluctuate greatly; the food waste of bakeries also fluctuates, because bakeries often overproduce to meet the needs of display, and the bread waste they generate can be reused or donated; in the subsequent collection and transportation, the collection and transportation points can make use of the characteristics of food waste generated by different types of catering units, and give priority to the collection and transportation of high-oil mixtures like those in hot pot restaurants. High oil itself has high utilization value but will cause blockage problems in food waste treatment facilities. Separate collection and transportation is conducive to reducing sorting costs. With the diversification of catering industry business models in the future, terminal enterprises can establish a dynamic list of collection objects and collection content, comprehensively evaluate the recycling value of food waste from various types of catering units, and improve the efficiency of resource utilization of food waste.

[0088] The generation of restaurant kitchen waste is highly correlated with raw materials. Taking Chinese restaurants, which generate the most restaurant kitchen waste, as an example, material flow analysis can provide a reference for the subsequent collection, transportation and disposal of restaurant kitchen waste. The results are as follows: Fig. 9As shown; through a survey of a Chinese restaurant, taking 100kg of raw materials as the benchmark, the main raw materials are meat, vegetables, staple foods, sauces, seafood, etc., among which staple foods include rice, noodles, etc., and sauces contain oil, salt, etc. Food waste can be divided into avoidable and unavoidable categories; inevitable waste refers to garbage that is thrown away due to quality needs or inedible during the cooking process, such as bones, shells, and burnt food, and avoidable waste refers to garbage that can be eaten but thrown away; the weight of avoidable garbage is calculated as the weight of remaining kitchen waste minus the weight of unavoidable garbage, and the weight of unavoidable garbage is based on the ratio of poultry meat to bones of 8:2, the ratio of livestock meat to bones of 7:3, and the ratio of seafood meat to shells (or bones) of 9:1; the analysis results show that 86.6% of the weight of raw materials is delivered to the table, 69.2% of the served portion is eaten, and the rest 13.4% is discarded as leftovers, and it is calculated that the ratio of avoidable and unavoidable waste is about 1:3; most of the food waste is protein, organic and lipids, among which meat and sauces with the highest fat content account for 31.5% of the waste. Extracting crude oil from them is the main source of profit for the subsequent resource utilization of food waste; the companies responsible for terminal disposal can communicate and coordinate with the food waste collection and transportation units according to the food waste feed needs, classify the sundries, food residues, catering waste oil, etc., improve the quality and efficiency of food waste collection and transportation, and choose sorting, crushing and extrusion pretreatment technology to effectively reduce the loss rate of organic materials to improve the resource utilization rate of food waste.

[0089] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A method for predicting the generation and discharge of urban kitchen waste based on BPN-RF, characterized in that: The following steps are involved: Step 1: Collect input data of number of seats, number of services, raw materials, turnover and output data of corresponding restaurant kitchen waste; Step 2: Use the data from step 1 to train the BPN network; use the assumed number of seats, number of services, raw materials, and turnover to predict the value of restaurant kitchen waste using the trained BPN network; Step 3: Perform initial interval division according to the frequency of the predicted kitchen waste value, and perform secondary interval division according to the initial interval division; Step 4: Using the secondary interval division and the corresponding number of seats, per capita restaurant kitchen waste, per capita catering consumption, and local per capita catering consumption as input data and the secondary interval division data as output data, the RF algorithm is trained; Step 5: Use the number of seats, per capita food waste, per capita catering consumption, and local per capita catering consumption to predict the trained RF algorithm and obtain the predicted value of the food waste interval.

2. The method for predicting the generation and discharge of urban kitchen waste based on BPN-RF according to claim 1 is characterized in that: The initial interval division is based on the frequency of occurrence of the kitchen waste value calculated in the first interval.

3. The method for predicting the generation and discharge of urban kitchen waste based on BPN-RF according to claim 2 is characterized in that: The initial interval division includes the first frequency zone, the second frequency zone and the third frequency zone.

4. The method for predicting the generation and discharge of urban kitchen waste based on BPN-RF according to claim 3 is characterized in that: The frequency of the first frequency region is greater than or equal to the first frequency threshold; the frequency of the third frequency region is less than the third frequency threshold; and the frequency of the second frequency region is between the first frequency threshold and the third frequency threshold.

5. The method for predicting the generation and discharge of urban kitchen waste based on BPN-RF according to claim 1 is characterized in that: The secondary interval division is to set the second interval when the cumulative distribution value of food waste after the initial interval division is less than the first food waste threshold.

6. The method for predicting the generation and discharge of urban restaurant kitchen waste based on BPN-RF according to claim 1, characterized in that: The secondary interval division is to set the third interval when the cumulative distribution value of the food waste after the initial interval division is greater than or equal to the first food waste threshold and less than or equal to the second food waste threshold.

7. The method for predicting the generation and discharge of urban kitchen waste based on BPN-RF according to claim 1 is characterized in that: The secondary interval division is to set the fourth interval when the cumulative distribution value of food waste after the initial interval division is greater than the second food waste threshold.

8. The method for predicting the generation and discharge of urban kitchen waste based on BPN-RF according to claim 1 is characterized in that: The assumed number of seats, number of services, raw materials and turnover are expanded by combining them in a Cartesian product.

9. The urban kitchen waste generation and discharge prediction system based on BPN-RF is characterized by: include: a memory for storing instructions executable by a processor; A processor, used to execute instructions to implement the method for predicting the generation and discharge of urban food waste based on BPN-RF as described in any one of claims 1 to 8.

10. A computer readable medium storing computer program code, characterized in that: When the computer program code is executed by a processor, the method for predicting the generation and discharge of urban food waste based on BPN-RF as described in any one of claims 1 to 8 is implemented.

Images