Capacity sales mode-based gas storage pricing method

Abstract

The invention discloses a gas storage pricing method based on a capacity sales mode, and the method comprises the steps: obtaining historical gas storage capacity sales data and historical influence factors as training data, carrying out the preprocessing of the training data, constructing a capacity sales prediction model, and training the capacity sales prediction model according to the preprocessed training data, obtaining a trained capacity sales prediction model; real-time influence factors are collected, and the real-time influence factors are used as the input of the trained capacity sales prediction model to predict the sales volume of the future gas storage capacity; and obtaining the total gas storage capacity, and obtaining the price in the capacity sales mode according to the future sales volume of the gas storage capacity and the total gas storage capacity. According to the invention, multi-level training is carried out on the capacity sales prediction model, so that the capacity sales prediction model can accurately predict the sales volume of the gas storage capacity, then pricing of the capacity sales price is carried out according to the predicted sales volume, workers can be assisted to adjust the pricing in time, and the working efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of gas storage, and particularly to a gas storage pricing method based on a capacity sales model. Background Art

[0002] The formulation of the gas storage fee in a gas storage reservoir follows the principles of allowing costs, reasonable profits, openness and transparency, and simplicity of operation. The service cost method is used to determine the basic gas storage rate of a joint-venture gas storage company. The transaction prices of specific gas storage service models and products are referenced to the basic gas storage rate and fluctuate within a certain range in combination with market supply and demand conditions.

[0003] The essence of the capacity sale model is a model that provides relatively long-term and stable warehousing services for customers. In terms of business model, it adopts a form of charging gas storage usage fees. Therefore, the price of capacity sale only has the gas storage rate, including the capacity rate and the usage rate. Under the capacity sale model, due to the exclusivity of capacity occupation, once the supply and demand parties sign a purchase contract, the sold capacity cannot be used by other users. Therefore, a fixed capacity fee is charged regardless of whether the customer uses it or not. At the same time, capacity sale is a warehousing service contract with a long cycle and a large capacity, and belongs to the user type that is preferentially guaranteed. Short-term warehousing services and interruptible service models can only be satisfied if there is a surplus after meeting the users of the capacity sale model. This poses some requirements for the operation and management of gas storage companies. On the one hand, once the capacity transaction is concluded, it is mandatory for the gas storage company to reserve the gas storage capacity agreed in the contract for the customer, which also has a strong restrictive effect on the use of other capacities of the gas storage company and increases the management difficulty of gas storage reservoir capacity allocation. On the other hand, the dispatching management should be concentrated to ensure its demand. Especially when there is a conflict in gas injection and production demands among different types of users during gas supply shortages, the gas storage company needs to coordinate all parties and adjust the capacity allocation and injection-production capacity allocation plan to meet the users of the capacity sale model, which means sacrificing the certain demands and interests of other users. In addition, under the capacity sale model, customers are also allowed to rent out the remaining unused capacity. Based on the above analysis, compared with the short-term warehousing service model, the capacity sale model should have a higher fixed capacity price as a constraint.

[0004] In the prior art, it is necessary to adjust the capacity price according to the capacity sales volume, usually adjusted manually based on historical data. However, manual adjustment not only takes time and effort but also has the problem of untimely adjustment. Summary of the Invention

[0005] The purpose of the present invention is to provide a gas storage pricing method based on a capacity sales model, which solves the problems existing in the prior art.

[0006] The present invention is achieved by the following technical solutions:

[0007] A gas storage pricing method based on a capacity sales model, comprising:

[0008] Obtain historical gas storage capacity sales data and historical influencing factors, and use the historical gas storage capacity sales data and historical influencing factors as training data;

[0009] Preprocess the training data to obtain preprocessed training data;

[0010] Construct a capacity sales prediction model, and train the capacity sales prediction model according to the preprocessed training data to obtain a trained capacity sales prediction model;

[0011] Collect real-time influencing factors, and use the real-time influencing factors as the input of the trained capacity sales prediction model to predict the future sales volume of gas storage capacity;

[0012] Obtain the total gas storage capacity, and based on the future sales volume of gas storage capacity and the total gas storage capacity, obtain the price under the capacity sales model to complete gas storage pricing.

[0013] In a possible real-time manner, the historical influencing factors include temperature data, season data, date type data, and user type percentage data;

[0014] The temperature data includes the highest temperature, average temperature, and lowest temperature at consecutive time points;

[0015] The season data includes spring, summer, autumn, and winter, and spring, summer, autumn, and winter are represented by different numbers respectively;

[0016] The date type data includes weekdays, non-weekdays, and holidays, and weekdays, non-weekdays, and holidays are represented by different numbers respectively;

[0017] The user type percentage data includes the percentage of residential users, the percentage of industrial users, and the percentage of commercial users.

[0018] In a possible real-time manner, preprocessing the training data to obtain preprocessed training data includes:

[0019] Detect the outlier data corresponding to the training data after missing value processing, and correct the outlier data to obtain the corrected training data;

[0020] Normalize the corrected training data to obtain the preprocessed training data.

[0021] In a possible real-time manner, a capacity sales prediction model is constructed and trained based on the preprocessed training data to obtain a trained capacity sales prediction model, including

[0022] Using a BP neural network as the capacity sales prediction model and constructing an error function corresponding to the capacity sales prediction model;

[0023] Initializing the weight parameters of the capacity sales prediction model to obtain multiple groups of weight parameters of the capacity sales prediction model;

[0024] According to the preprocessed training data and the error function, updating the weight parameters of multiple groups of capacity sales prediction models once to obtain multiple groups of weight parameters after the first update;

[0025] Selecting the optimal group of weight parameters from the multiple groups of weight parameters after the first update, and performing a second update on the other groups of weight parameters according to the optimal group of weight parameters to obtain multiple groups of weight parameters after the second update;

[0026] According to the multiple groups of weight parameters after the second update, reselecting the optimal group of weight parameters and performing a jump update on the optimal group of weight parameters to obtain the updated optimal group of weight parameters to jump out of the local optimum;

[0027] After repeating the training multiple times for all weight parameters, determining the loss function value corresponding to the optimal group of weight values, and judging whether the loss function value is less than the set threshold. If so, the training is completed, and the optimal group of weight values is used as the weight values of the capacity sales prediction model; otherwise, the capacity sales prediction model is trained in the next round.

[0028] In a possible real-time manner, the error function corresponding to the capacity sales prediction model is constructed as:

[0029]

[0030] where E represents the error function, d j represents the actual output corresponding to the j-th neuron in the output layer of the capacity sales prediction model, y j represents the expected output corresponding to the j-th neuron in the output layer of the capacity sales prediction model, j = 1, 2,..., n, and n represents the total number of neurons in the output layer.

[0031] In a possible real-time manner, initializing the weight parameters of the capacity sales prediction model to obtain multiple groups of weight parameters of the capacity sales prediction model, including:

[0032] Determining the weight initialization rule as:

[0033] wi = w u - rand·(w u - w l )

[0034] where w i represents the weight after initialization, w u represents the upper limit of the weight value, w l represents the lower limit of the weight value, and rand represents a random number between 0 and 1;

[0035] According to the weight initialization rule, each weight in the capacity sales prediction model is initialized N times to obtain N sets of weight parameters.

[0036] In a possible real-time manner, based on the preprocessed training data and the error function, the weight parameters of multiple capacity sales prediction models are updated once to obtain multiple sets of updated weight parameters, including:

[0037] Randomly select a set of weight parameters from multiple sets of weight parameters of the capacity sales prediction model to obtain the first target weight parameter, and apply the first target weight parameter to the capacity sales prediction model;

[0038] Use the historical influencing factors in the preprocessed training data as the input of the capacity sales prediction model to obtain the actual output, use the historical gas storage capacity sales data as the expected output, and obtain the first error function value corresponding to the first target weight parameter according to the actual output, expected output, and error function;

[0039] Update the first target weight parameter to obtain the updated first target weight parameter, and the update is as follows:

[0040]

[0041] where represents the m-th weight in the first target weight parameter before update, represents the m-th weight in the updated first target weight parameter, represents the m-th weight in the first target weight parameter of the previous update, and rand represents a random number between 0 and 1;

[0042] Use the historical influencing factors in the preprocessed training data as the input of the capacity sales prediction model to obtain the actual output, use the historical gas storage capacity sales data as the expected output, and obtain the second error function value corresponding to the updated first target weight parameter according to the actual output, expected output, and error function;

[0043] Determine whether the second error function value is less than the first error function value. If so, accept the current update of the first target weight parameter; otherwise, do not accept the current update of the first target weight parameter.

[0044] Determine whether the weight parameters of all groups have been updated. If so, obtain the multi-group weight parameters after one update; otherwise, return to the step of obtaining the first target weight parameter.

[0045] In a possible real-time manner, select the optimal group of weight parameters from the multi-group weight parameters after one update, and perform a secondary update on the weight parameters of other groups according to the optimal group of weight parameters to obtain the multi-group weight parameters after the secondary update, including:

[0046] Select the optimal group of weight parameters from the multi-group weight parameters after one update;

[0047] Randomly select a group of weight parameters from the remaining weight parameters to obtain the second target weight parameter, and apply the second target weight parameter to the capacity sales prediction model;

[0048] Use the historical influencing factors in the preprocessed training data as the input of the capacity sales prediction model to obtain the actual output, use the historical gas storage capacity sales data as the expected output, and obtain the third error function value according to the actual output, the expected output, and the error function;

[0049] Update the second target weight parameter according to the optimal group of weight parameters to obtain the updated second target weight parameter. The update is as follows:

[0050]

[0051]

[0052]

[0053] Where, represents the m-th weight in the second target weight parameter before update, represents the m-th weight in the second target weight parameter after update, α represents the first update parameter, represents element-wise multiplication, L(β) represents the first intermediate parameter, β represents the second update parameter, α0 represents the third update parameter, β = 1.48, α0 = 0.011, represents the m-th weight in the optimal group of weight parameters, μ represents the fourth update parameter, and μ follows a normal distribution δ μ represents the second intermediate parameter, Γ() represents the gamma function, π represents pi, V represents the fifth update parameter, and V follows a normal distribution δ V represents the third intermediate parameter, δ V = 1;

[0054] Using the historical influencing factors in the preprocessed training data as the input of the capacity sales prediction model to obtain the actual output, using the historical gas storage capacity sales data as the expected output, and according to the actual output, the expected output, and the error function, obtaining the fourth error function value corresponding to the updated second target weight parameter;

[0055] Determine whether the fourth error function value is less than the third error function value. If so, accept this update of the second target weight parameter; otherwise, do not accept this update of the second target weight parameter;

[0056] Determine whether all groups of weight parameters after one update have been updated. If so, obtain the groups of weight parameters after the second update; otherwise, return to the step of obtaining the second target weight parameter.

[0057] In a possible real-time manner, according to the groups of weight parameters after the second update, reselect the optimal group of weight parameters and perform a jump update on the optimal group of weight parameters to obtain the updated optimal group of weight parameters, including:

[0058] Obtain the error function values corresponding to the groups of weight parameters after the second update, and determine the weight parameters after the second update with the smallest error function value as the optimal group of weight parameters;

[0059] Perform a jump update on the optimal group of weight parameters to obtain the updated optimal group of weight parameters. The update is as follows:

[0060]

[0061] where, represents the m-th weight in the optimal group of weight parameters before the update, represents the m-th weight in the optimal group of weight parameters after the update, r represents a random number between -0.8 and +0.8, λ represents a scaling factor, λ = 0.1(w u - w l ), w u represents the upper limit of the weight value, w l represents the lower limit of the weight value.

[0062] In a possible real-time manner, obtain the total gas storage capacity, and according to the future sales volume of the gas storage capacity and the total gas storage capacity, obtain the price in the capacity sales mode, including:

[0063] Obtain the total gas storage capacity, divide the future sales volume of the gas storage capacity by the total gas storage capacity to obtain the proportion of the future sales volume of the gas storage capacity;

[0064] Judge whether the sales volume ratio is less than the set threshold. If so, obtain the floating capacity rate as 10 / 9 * base capacity rate * sales volume ratio according to the sales volume ratio; otherwise, obtain the floating capacity rate as 5 / 3 * base capacity rate * (1 - sales volume ratio) according to the sales volume ratio, where the base capacity rate is a constant.

[0065] Obtain the fixed capacity rate as the base capacity rate + floating capacity rate according to the floating capacity rate, and obtain the capacity sales price in the capacity sales mode as the fixed capacity rate + usage rate according to the fixed capacity rate. Both the base capacity rate and the usage rate are constants.

[0066] A gas storage pricing method based on the capacity sales mode provided by the present invention constructs a capacity sales prediction model, predicts the future sales volume of the gas storage capacity through the capacity sales prediction model, and then prices the capacity sales price according to the predicted sales volume, which not only solves the problem of time-consuming and laborious manual pricing, but also can assist the staff to adjust the price in time and improve work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following will briefly introduce the drawings required in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings. In the drawings:

[0068] Figure 1 It is a flowchart of a gas storage pricing method based on the capacity sales mode provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0069] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in combination with the embodiments and the drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0070] Embodiment 1

[0071] As Figure 1 shown, this embodiment provides a gas storage pricing method based on the capacity sales mode, including:

[0072] S11. Obtain historical gas storage capacity sales data and historical influencing factors, and use the historical gas storage capacity sales data and historical influencing factors as training data.

[0073] When obtaining historical gas storage capacity sales data and historical influencing factors, it can be based on time points. At each time point, historical gas storage capacity sales data and historical influencing factors are obtained, and one time point can represent one day. Therefore, historical data can be collected on a daily basis. Historical influencing factors refer to the factors that affect gas storage capacity sales. For example, the merchant proportion and temperature level will affect the overall gas consumption volume, thereby affecting the purchased gas storage capacity. Historical gas storage capacity sales data refers to the corresponding gas storage capacity sales volume under the action of historical influencing factors.

[0074] S12. Preprocess the training data to obtain the preprocessed training data.

[0075] The preprocessing can include missing value processing, outlier processing, and normalization processing. The purpose of missing value processing and outlier processing is to make the data more regular, so that more accurate training can be carried out. The purpose of normalization is to normalize the data values into a certain interval, which not only conforms to the data input type of the gas storage capacity sales prediction model, but also can effectively reduce the data processing volume.

[0076] S13. Construct a capacity sales prediction model, and train the capacity sales prediction model according to the preprocessed training data to obtain the trained capacity sales prediction model.

[0077] When training the capacity sales prediction model, it can be first updated once according to the weight parameter values to generate better values, and then based on the optimal weight parameters, the other weight parameters are updated twice to further optimize the weight parameters. Then, the optimal weight parameter among all the weight parameters is selected again, and the optimal weight parameter is updated by jumping, so that the weight parameters can jump out of the local optimal value, avoiding the problem of local optimal value caused by the traditional use of the gradient descent method for updating.

[0078] S14. Collect real-time influencing factors, and use the real-time influencing factors as the input of the trained capacity sales prediction model to predict the future sales volume of gas storage capacity.

[0079] S15. Obtain the total gas storage capacity, and based on the future sales volume of gas storage capacity and the total gas storage capacity, obtain the price under the capacity sales mode to complete gas storage pricing.

[0080] A gas storage pricing method based on the capacity sales mode provided by the present invention constructs a capacity sales prediction model, predicts the future sales volume of gas storage capacity through the capacity sales prediction model, and then prices the capacity sale price according to the predicted sales volume, which not only solves the problem of time-consuming and laborious manual pricing, but also can assist the staff to adjust the pricing in time and improve work efficiency.

[0081] In a possible real-time manner, the historical influencing factors include temperature data, season data, date type data, and user type percentage data; the temperature data includes the highest temperature, average temperature, and lowest temperature at consecutive time points; the season data includes spring, summer, autumn, and winter, and spring, summer, autumn, and winter are represented by different numbers respectively. For example, spring, summer, autumn, and winter can be represented by 1, 2, 3, and 4 respectively; the date type data includes working days, non-working days, and holidays, and working days, non-working days, and holidays are represented by different numbers respectively. For example, working days can be represented by 1 to 5 according to Monday to Friday; non-working days can be represented by 6 and 7 according to Saturday to Sunday; holidays are uniformly represented by 8. It should be noted that each time point can only be one date type, and holidays take precedence. For example, when the time point is a holiday, it cannot be a working day or a non-working day; temperature, season, and date type will all cause changes in the user's gas demand, and the total storage capacity is fixed. As the gas demand changes, the user's demand for capacity purchase is also changing. Therefore, a neural network model can be used to simulate this non-linear change to predict the capacity sales volume. The user type percentage data includes the percentage of residential users, the percentage of industrial users, and the percentage of commercial users. Different users naturally have different demands for purchasing capacity. Therefore, the user proportion needs to be considered.

[0082] Optionally, when collecting real-time influencing factors, the temperature data can be sourced from the weather forecast of the meteorological bureau, and other data can be collected normally.

[0083] In a possible real-time manner, the training data is preprocessed to obtain the preprocessed training data, including: detecting the outlier data corresponding to the training data after missing value processing, and correcting the outlier data to obtain the corrected training data; normalizing the corrected training data to obtain the preprocessed training data.

[0084] In step S13, a capacity sales prediction model is constructed, and the capacity sales prediction model is trained according to the preprocessed training data to obtain the trained capacity sales prediction model, including

[0085] S131. Use a BP neural network as the capacity sales prediction model and construct an error function corresponding to the capacity sales prediction model.

[0086] S132. Initialize the weight parameters of the capacity sales prediction model to obtain multiple groups of weight parameters of the capacity sales prediction model.

[0087] The weights of each part in the capacity sales prediction model can be initialized multiple times in a random initialization manner to obtain multiple groups of weight parameters of the capacity sales prediction model.

[0088] S133. Update the weight parameters of multiple capacity sales prediction models once according to the preprocessed training data and the error function, and obtain multiple groups of weight parameters after the first update.

[0089] By updating the weight parameters of multiple capacity sales prediction models once, the multiple groups of weight parameters can be updated in a more optimal direction.

[0090] S134. Select the optimal group of weight parameters from the multiple groups of weight parameters after the first update, and perform a second update on the other groups of weight parameters according to the optimal group of weight parameters to obtain multiple groups of weight parameters after the second update.

[0091] After the weight parameters of multiple capacity sales prediction models are updated once, the other weight parameters are guided by the optimal weight parameters for updating, so as to generate weight parameters with a smaller loss function value.

[0092] S135. According to the multiple groups of weight parameters after the second update, reselect the optimal group of weight parameters, and perform a jump update on the optimal group of weight parameters to obtain the updated optimal group of weight parameters to jump out of the local optimum.

[0093] After the weight parameters are updated, the optimal weight parameter value may fall into the local optimum. Perform a jump update on it to generate a perturbation, so that the optimal weight parameter vibrates, thus jumping out of the local optimum.

[0094] S136. After repeating the training for multiple times on all weight parameters, determine the loss function value corresponding to the optimal group of weight values, and judge whether the loss function value is less than the set threshold. If so, the training is completed, and the optimal group of weight values is used as the weight values of the capacity sales prediction model; otherwise, the capacity sales prediction model is trained in the next round.

[0095] First, perform multiple rounds of training on the weight parameters to complete the basic update of the weight parameters, and then obtain the error function value corresponding to each group of weight parameters. The smaller the error function value, the better the weight parameters. When there are qualified weight parameters, the training can be completed; otherwise, continue to update, so as to obtain a capacity sales prediction model with better prediction effect.

[0096] It should be noted that the main purpose of the training method in this embodiment is to train the optimal weights of the neural network. Other parameters of the capacity sales prediction model can be trained according to this method, or other parameters of the capacity sales prediction model can be trained according to the conventional method.

[0097] In step S131, the error function corresponding to the capacity sales prediction model is constructed as:

[0098]

[0099] Among them, E represents the error function, and d j represents the actual output corresponding to the j-th neuron in the output layer of the capacity sales prediction model, and y j represents the expected output corresponding to the j-th neuron in the output layer of the capacity sales prediction model, where j = 1, 2,..., n, and n represents the total number of neurons in the output layer.

[0100] In step S132, the weight parameters of the capacity sales prediction model are initialized to obtain multiple groups of weight parameters of the capacity sales prediction model, including:

[0101] Determine the weight initialization rule as:

[0102] w i = w u - rand·(w u - w l )

[0103] Among them, w i represents the initialized weight, w u represents the upper limit of the weight value, w l represents the lower limit of the weight value, and rand represents a random number between 0 and 1.

[0104] According to the weight initialization rule, each weight in the capacity sales prediction model is initialized N times to obtain N groups of weight parameters.

[0105] In step S133, based on the preprocessed training data and the error function, the multiple groups of weight parameters of the capacity sales prediction model are updated once to obtain multiple groups of weight parameters after one update, including:

[0106] S1331. Randomly select a group of weight parameters from the multiple groups of weight parameters of the capacity sales prediction model to obtain the first target weight parameter, and apply the first target weight parameter to the capacity sales prediction model.

[0107] S1332. Use the historical influencing factors in the preprocessed training data as the input of the capacity sales prediction model to obtain the actual output, use the historical gas storage capacity sales data as the expected output, and obtain the first error function value corresponding to the first target weight parameter according to the actual output, expected output, and error function.

[0108] S1333. Update the first target weight parameter to obtain the updated first target weight parameter, and the update is as follows:

[0109]

[0110] Among them, represents the m-th weight in the first target weight parameter before update, represents the m-th weight in the first target weight parameter after update, represents the m-th weight in the first target weight parameter of the previous update, and rand represents a random number between 0 and 1. It should be noted that when the first update is performed, it does not exist and can be set to 0.

[0111] S1334. Use the historical influencing factors in the preprocessed training data as the input of the capacity sales prediction model to obtain the actual output, use the historical gas storage capacity sales data as the expected output, and obtain the second error function value corresponding to the updated first target weight parameter according to the actual output, the expected output, and the error function.

[0112] S1335. Determine whether the second error function value is less than the first error function value. If so, accept this update of the first target weight parameter; otherwise, do not accept this update of the first target weight parameter.

[0113] S1336. Determine whether the weight parameters of all groups have been updated. If so, obtain the multi-group weight parameters after one update; otherwise, return to the step of obtaining the first target weight parameter.

[0114] In step S134, select the optimal group of weight parameters from the multi-group weight parameters after one update, and perform a secondary update on the weight parameters of other groups according to the optimal group of weight parameters to obtain the multi-group weight parameters after secondary update, including:

[0115] S1341. Select the optimal group of weight parameters from the multi-group weight parameters after one update.

[0116] S1342. Randomly select a group of weight parameters from the remaining weight parameters to obtain the second target weight parameter, and apply the second target weight parameter to the capacity sales prediction model.

[0117] S1343. Use the historical influencing factors in the preprocessed training data as the input of the capacity sales prediction model to obtain the actual output, use the historical gas storage capacity sales data as the expected output, and obtain the third error function value according to the actual output, the expected output, and the error function.

[0118] S1344. Update the second target weight parameter according to the optimal group of weight parameters to obtain the updated second target weight parameter. The update is as follows:

[0119]

[0120]

[0121]

[0122] Among them, represents the m-th weight in the second target weight parameter before update, represents the m-th weight in the second target weight parameter after update, α represents the first update parameter, represents element-wise multiplication, L(β) represents the first intermediate parameter, β represents the second update parameter, α0 represents the third update parameter, β = 1.48, α0 = 0.011, represents the m-th weight in the optimal set of weight parameters, μ represents the fourth update parameter, and μ follows a normal distribution δ μ represents the second intermediate parameter, Γ() represents the gamma function, π represents pi, V represents the fifth update parameter, and V follows a normal distribution δ V represents the third intermediate parameter, δ V = 1.

[0123] S1345. Use the historical influencing factors in the preprocessed training data as the input of the capacity sales prediction model to obtain the actual output, use the historical gas storage capacity sales data as the expected output, and obtain the fourth error function value corresponding to the updated second target weight parameter according to the actual output, the expected output, and the error function.

[0124] S1346. Determine whether the fourth error function value is less than the third error function value. If so, accept this update of the second target weight parameter; otherwise, do not accept this update of the second target weight parameter.

[0125] S1347. Determine whether all groups of weight parameters after one update have been updated. If so, obtain the groups of weight parameters after the second update; otherwise, return to the step of obtaining the second target weight parameter.

[0126] In step S135, according to the groups of weight parameters after the second update, reselect the optimal group of weight parameters and perform a jump update on the optimal group of weight parameters to obtain the updated optimal group of weight parameters, including:

[0127] S1351. Obtain the error function values corresponding to the groups of weight parameters after the second update, and determine the weight parameters after the second update with the smallest error function value as the optimal group of weight parameters.

[0128] S1352. Perform a jump update on the optimal group of weight parameters to obtain the updated optimal group of weight parameters. The update is as follows:

[0129]

[0130] Among them, represents the m-th weight in the optimal set of weight parameters before update, represents the m-th weight in the optimal set of weight parameters after update, r represents a random number between -0.8 and +0.8, λ represents a scaling factor, and λ = 0.1(w u - w l ), w u represents the upper limit of the weight value, and w l represents the lower limit of the weight value.

[0131] In step S15, obtain the total gas storage capacity. According to the future sales volume of the gas storage capacity and the total gas storage capacity, obtain the price in the capacity sales mode, including:

[0132] S151. Obtain the total gas storage capacity. Divide the future sales volume of the gas storage capacity by the total gas storage capacity to obtain the proportion of the future sales volume of the gas storage capacity. The total gas storage capacity can be data pre-stored in the database and is only used here.

[0133] S152. Determine whether the proportion of the sales volume is less than a set threshold (for example, the threshold can be 0.6). If so, obtain the floating capacity rate as 10 / 9 * basic capacity rate * proportion of the sales volume according to the proportion of the sales volume; otherwise, obtain the floating capacity rate as 5 / 3 * basic capacity rate * (1 - proportion of the sales volume), where the basic capacity rate is a constant.

[0134] The basic capacity rate represents the basic price required to purchase a unit of gas storage capacity, which is a fixed constant, while the floating capacity rate represents the floating price required to purchase a unit of gas storage capacity. The sum of the basic price and the floating price is the price of selling a unit of capacity (i.e., the fixed capacity rate). The usage fee represents the maintenance fee required to use a unit of gas storage capacity, which can also be a fixed constant.

[0135] Therefore, to determine the final capacity sales price, it is necessary to determine the fixed capacity rate and the usage fee, which is equivalent to determining the floating capacity rate. The floating capacity rate can be obtained by predicting the proportion of capacity purchase, so as to achieve faster response.

[0136] S153. Obtain the fixed capacity rate as the basic capacity rate + floating capacity rate according to the floating capacity rate, and obtain the capacity sales price in the capacity sales mode as the fixed capacity rate + usage rate according to the fixed capacity rate. Both the basic capacity rate and the usage rate are constants.

[0137] A gas storage pricing method based on a capacity sales model provided by the present invention trains a capacity sales prediction model at multiple levels, enabling the capacity sales prediction model to accurately predict the capacity sales volume in future periods. Then, the final capacity sales price is obtained based on the capacity sales volume, making the pricing adjustment more timely. Compared with manually analyzing historical data, it has higher efficiency and can achieve more accurate pricing.

[0138] Those skilled in the art should understand that the embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented in various computer languages. For example, object-oriented programming languages such as Java and interpreted scripting languages such as JavaScript.

[0139] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to generate a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more of the processes or multiple processes and / or blocks Figure 1 one or more of the blocks or multiple blocks.

[0140] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one or more of the processes or multiple processes and / or blocks Figure 1 one or more of the blocks or multiple blocks.

[0141] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide means for implementing the functions specified in Figure 1 one or more of the processes or multiple processes and / or blocksFigure 1 Steps of functions specified in one or more boxes.

[0142] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications to these embodiments once they learn the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0143] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A gas storage pricing method based on a capacity sales model, characterized in that Including: Obtain historical gas storage capacity sales data and historical influencing factors, and use the historical gas storage capacity sales data and historical influencing factors as training data; Preprocess the training data to obtain preprocessed training data; Construct a capacity sales prediction model, and train the capacity sales prediction model according to the preprocessed training data to obtain a trained capacity sales prediction model; Collect real-time influencing factors, and use the real-time influencing factors as the input of the trained capacity sales prediction model to predict the future sales volume of gas storage capacity; Obtain the total gas storage capacity, and based on the future sales volume of gas storage capacity and the total gas storage capacity, obtain the price under the capacity sales mode to complete gas storage pricing.

2. The gas storage pricing method based on the capacity sales model according to claim 1, wherein The historical influencing factors include temperature data, season data, date type data, and user type percentage data; The temperature data includes the highest temperature, average temperature, and lowest temperature at consecutive time points; The season data includes spring, summer, autumn, and winter, and spring, summer, autumn, and winter are represented by different numbers respectively; The date type data includes working days, non-working days, and holidays, and working days, non-working days, and holidays are represented by different numbers respectively; The user type percentage data includes the percentage of residential users, the percentage of industrial users, and the percentage of commercial users.

3. The gas storage pricing method based on the capacity sales model according to claim 1, characterized in that Preprocessing the training data to obtain preprocessed training data includes: Detect the outlier data corresponding to the training data after missing value processing, and correct the outlier data to obtain the corrected training data; Normalize the corrected training data to obtain the preprocessed training data.

4. The gas storage pricing method based on the capacity sales model according to claim 2, wherein Construct a capacity sales prediction model, and train the capacity sales prediction model according to the preprocessed training data to obtain a trained capacity sales prediction model, including Use a BP neural network as the capacity sales prediction model, and construct an error function corresponding to the capacity sales prediction model; Initialize the weight parameters of the capacity sales prediction model to obtain multiple groups of weight parameters of the capacity sales prediction model; According to the preprocessed training data and the error function, update the weight parameters of multiple groups of capacity sales prediction models once to obtain multiple groups of weight parameters after the first update; Select the optimal group of weight parameters from the multiple groups of weight parameters after the first update, and perform a second update on the other groups of weight parameters according to the optimal group of weight parameters to obtain multiple groups of weight parameters after the second update; According to the multiple groups of weight parameters after the second update, reselect the optimal group of weight parameters, and perform a jump update on the optimal group of weight parameters to obtain the updated optimal group of weight parameters to jump out of the local optimum; After repeating the training multiple times for all weight parameters, determine the loss function value corresponding to the optimal group of weight values, and judge whether the loss function value is less than the set threshold. If so, the training is completed, and the optimal group of weight values is used as the weight values of the capacity sales prediction model. Otherwise, perform the next round of training on the capacity sales prediction model.

5. The gas storage pricing method based on the capacity sales model according to claim 4, characterized in that Construct the error function corresponding to the capacity sales prediction model as: where E represents the error function, d j represents the actual output corresponding to the j-th neuron in the output layer of the capacity sales prediction model, y j represents the expected output corresponding to the j-th neuron in the output layer of the capacity sales prediction model, j = 1, 2, …, n, and n represents the total number of neurons in the output layer.

6. The gas storage pricing method based on the capacity sales model according to claim 5, wherein Initialize the weight parameters of the capacity sales prediction model to obtain multiple groups of weight parameters of the capacity sales prediction model, including: Determine the weight initialization rule as: w i = w u - rand·(w u - w l ) Among them, w i represents the weight after initialization, w u represents the upper limit of the weight value, w l represents the lower limit of the weight value, and rand represents a random number between 0 and 1; According to the weight initialization rule, initialize each weight in the capacity sales prediction model N times to obtain N groups of weight parameters.

7. The gas storage pricing method based on the capacity sales model according to claim 5, characterized in that, Based on the preprocessed training data and the error function, update the multiple groups of weight parameters of the capacity sales prediction model once to obtain the multiple groups of weight parameters after one update, including: Randomly select a group of weight parameters from the multiple groups of weight parameters of the capacity sales prediction model to obtain the first target weight parameter, and apply the first target weight parameter to the capacity sales prediction model; Use the historical influencing factors in the preprocessed training data as the input of the capacity sales prediction model to obtain the actual output, use the historical gas storage capacity sales data as the expected output, and obtain the first error function value corresponding to the first target weight parameter according to the actual output, the expected output, and the error function; Update the first target weight parameter to obtain the updated first target weight parameter, and the update is: Among them, represents the m-th weight in the first target weight parameter before update, represents the m-th weight in the first target weight parameter after update, represents the m-th weight in the first target weight parameter of the previous update, and rand represents a random number between 0 and 1; Use the historical influencing factors in the preprocessed training data as the input of the capacity sales prediction model to obtain the actual output, use the historical gas storage capacity sales data as the expected output, and obtain the second error function value corresponding to the updated first target weight parameter according to the actual output, the expected output, and the error function; Judge whether the second error function value is less than the first error function value. If so, accept this update of the first target weight parameter; otherwise, do not accept this update of the first target weight parameter; Judge whether all groups of weight parameters have been updated. If so, obtain the multiple groups of weight parameters after one update; otherwise, return to the step of obtaining the first target weight parameter.

8. The gas storage pricing method based on the capacity sales model according to claim 5, characterized in that Select the optimal group of weight parameters from the multiple groups of weight parameters after one update, and perform a secondary update on the other groups of weight parameters according to the optimal group of weight parameters to obtain the multiple groups of weight parameters after the secondary update, including: Select the optimal group of weight parameters from the multiple groups of weight parameters after one update; Randomly select a group of weight parameters from the remaining weight parameters to obtain the second target weight parameter, and apply the second target weight parameter to the capacity sales prediction model; Use the historical influencing factors in the preprocessed training data as the input of the capacity sales prediction model to obtain the actual output, use the historical gas storage capacity sales data as the expected output, and obtain the third error function value according to the actual output, the expected output, and the error function; Update the second target weight parameter according to the optimal group of weight parameters to obtain the updated second target weight parameter, and the update is: Among them, represents the m-th weight in the second target weight parameter before update, represents the m-th weight in the second target weight parameter after update, α represents the first update parameter, represents pointwise multiplication, L(β) represents the first intermediate parameter, β represents the second update parameter, α0 represents the third update parameter, β = 1.48, α0 = 0.011, represents the m-th weight in the optimal set of weight parameters, μ represents the fourth update parameter, and μ follows a normal distribution δ μ represents the second intermediate parameter, Γ() represents the gamma function, π represents pi, V represents the fifth update parameter, and V follows a normal distribution δ V represents the third intermediate parameter, δ V = 1; Use the historical influencing factors in the preprocessed training data as the input of the capacity sales prediction model to obtain the actual output, use the historical gas storage capacity sales data as the expected output, and obtain the fourth error function value corresponding to the updated second target weight parameter according to the actual output, the expected output, and the error function; Determine whether the value of the fourth error function is less than the value of the third error function. If so, accept the current update of the second target weight parameter; otherwise, do not accept the current update of the second target weight parameter. Determine whether all groups of weight parameters after one update have been updated. If so, obtain the groups of weight parameters after the second update; otherwise, return to the step of obtaining the second target weight parameter.

9. The gas storage pricing method based on the capacity sales model according to claim 5, characterized in that According to the groups of weight parameters after the second update, re-select the optimal group of weight parameters and perform a jump update on the optimal group of weight parameters to obtain the updated optimal group of weight parameters, including: Obtain the error function values corresponding to the groups of weight parameters after the second update, and determine the weight parameter after the second update with the smallest error function value as the optimal group of weight parameters. Perform a jump update on the optimal group of weight parameters to obtain the updated optimal group of weight parameters. This update is as follows: Among them, represents the m-th weight in the optimal set of weight parameters before update, represents the m-th weight in the optimal set of weight parameters after update, r represents a random number between -0.8 and +0.8, λ represents a scaling factor, λ = 0.1(w u - w l ), w u represents the upper limit of the weight value, w l represents the lower limit of the weight value.

10. The gas storage pricing method based on the capacity sales model according to claim 1, wherein, Obtain the total gas storage capacity. According to the future sales volume of the gas storage capacity and the total gas storage capacity, obtain the price in the capacity sales mode, including: Obtain the total gas storage capacity. Divide the future sales volume of the gas storage capacity by the total gas storage capacity to obtain the proportion of the future sales volume of the gas storage capacity. Determine whether the proportion of the sales volume is less than the set threshold. If so, obtain the floating capacity rate as 10 / 9 * basic capacity rate * proportion of the sales volume according to the proportion of the sales volume; otherwise, obtain the floating capacity rate as 5 / 3 * basic capacity rate * (1 - proportion of the sales volume) according to the proportion of the sales volume, where the basic capacity rate is a constant. Obtain the fixed capacity rate as the basic capacity rate + floating capacity rate according to the floating capacity rate, and obtain the capacity sales price in the capacity sales mode as the fixed capacity rate + usage rate according to the fixed capacity rate. Both the basic capacity rate and the usage rate are constants.

Images