A power consumption prediction method based on a multi-task interpolation echo state network

Abstract

The application provides a power consumption prediction method based on a multi-task interpolation echo state network. It relates to the technical field of power consumption prediction and comprises the following steps: obtaining sewage data to be measured; obtaining input data according to the sewage data to be measured by using a synthetic time series; inputting the input data into a multi-task interpolation echo state network model to obtain a power consumption prediction value, wherein the power consumption prediction value is a pumping energy consumption value and an aeration energy consumption value, and the multi-task interpolation echo state network model comprises an input layer, a reservoir and an output layer. The application solves the problem that the pumping energy consumption and the aeration energy consumption in the sewage treatment process are difficult to accurately and efficiently predict in the prior art.

Description

Technical Field

[0001] This invention relates to the field of power consumption prediction technology, and in particular to a power consumption prediction method based on a multi-task interpolation echo state network. Background Technology

[0002] my country prioritizes ecological environmental protection and energy conservation and carbon reduction, leading to the widespread development of urban wastewater treatment plants. However, these plants face challenges such as high energy consumption and operating costs. The electricity consumption of wastewater treatment plants mainly consists of aeration energy consumption and pumping energy consumption. Pumping energy consumption is primarily caused by influent lift pumps and sludge return pumps, while aeration energy consumption refers to the electricity consumed during the process of introducing dissolved oxygen into the aeration tank using blowers. Accurate prediction of pumping and aeration energy consumption enables wastewater treatment plants to take targeted measures in either the pumping or aeration stages, thereby reducing energy consumption. This provides a basis for optimizing and controlling wastewater treatment processes, promoting the efficient and stable operation of wastewater treatment plants and sustainable ecological protection. Currently, the energy consumption of wastewater treatment plants is influenced by various biological and environmental factors, making the establishment of predictive models complex and challenging. Accurate and efficient prediction of pumping and aeration energy consumption in the wastewater treatment process remains difficult. Summary of the Invention

[0003] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a power consumption prediction method based on a multi-task interpolation echo state network. This invention solves the problem that it is difficult to accurately and efficiently predict the pumping energy consumption and aeration energy consumption in the sewage treatment process in existing technologies.

[0004] To achieve the above objectives, the present invention provides the following solution:

[0005] A power consumption prediction method based on a multi-task interpolation echo state network includes:

[0006] Acquire wastewater data;

[0007] Using a synthetic time series, the input data is obtained from the wastewater data to be tested;

[0008] The input data is fed into a multi-task interpolation echo state network model to obtain predicted power consumption values, wherein the predicted power consumption values ​​are pumping energy consumption values ​​and aeration energy consumption values. The multi-task interpolation echo state network model includes an input layer, a storage tank, and an output layer.

[0009] Preferably, the wastewater data to be tested includes:

[0010] Effluent COD concentration, effluent total phosphorus concentration, influent pH, effluent pH, DO concentration.

[0011] Preferably, acquiring the wastewater data to be tested includes:

[0012] Retrieve labeled and unlabeled data;

[0013] An interpolation method based on distance coefficients is used to obtain virtual interpolation points using unlabeled data;

[0014] The wastewater data to be tested is obtained based on the virtual interpolation points and the actual interpolation points corresponding to the labeled data.

[0015] Preferably, the distance coefficient-based interpolation method, which uses unlabeled data to obtain virtual interpolation points, includes:

[0016] Determine the label data at the first time step and the label data at the second time step;

[0017] The average value of the unlabeled data between the first and second time points is obtained based on the label data at the first time point and the label data at the second time point.

[0018] The virtual interpolation point is determined based on the average value of the unlabeled data.

[0019] Preferably, the step of obtaining the input data based on the wastewater data to be tested using a synthetic time series includes:

[0020] The labeled time series is obtained based on the actual interpolation points corresponding to the labeled data;

[0021] A virtual time series is obtained based on the virtual interpolation points;

[0022] A synthesized time series is obtained based on the labeled time series and the virtual time series;

[0023] The input data is obtained based on the synthesized time series.

[0024] Preferably, the network structure of the multi-task interpolation echo state network model is 5 - (k1N + k2N) - 2, where N is an initial positive integer, typically 10, and k... j ≥1 and a positive integer, representing the number of sub-reservoir pools possessed by each ESN module, k j N represents the size of the storage tank for the pumping energy consumption of Task 1 and the aeration energy consumption of Task 2, respectively, where j = 1, 2.

[0025] Preferably, it further includes:

[0026] Obtain the input data and predicted power consumption values;

[0027] The Laida criterion is used to detect outliers in the input data and the predicted power consumption. If outliers are found, they are replaced.

[0028] Preferably, the expression for the model of the multi-task interpolation echo state network is:

[0029]

[0030] Where s represents the original data. For the normalized data, s min and s max These are the minimum and maximum values ​​of the input data before normalization, respectively.

[0031] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0032] This invention provides a power consumption prediction method based on a multi-task interpolation echo state network, comprising: acquiring wastewater data to be measured; obtaining input data based on the wastewater data using a synthetic time series; inputting the input data into a multi-task interpolation echo state network model to obtain predicted power consumption values, wherein the predicted power consumption values ​​are pumping energy consumption values ​​and aeration energy consumption values, and the multi-task interpolation echo state network model includes an input layer, a storage tank, and an output layer. This invention achieves simultaneous and accurate prediction of pumping energy consumption and aeration energy consumption by preprocessing multi-rate sampling data and connecting parallel storage tanks, which is beneficial for ensuring optimized operation and energy conservation in wastewater treatment plants. Addressing the multi-rate phenomenon in the data collected from wastewater treatment plants, this invention designs an interpolation method based on distance coefficients to construct virtual points, enriching the data sample. Secondly, to address the problem of neglecting the correlation between prediction tasks of different energy consumption parameters, a multi-task learning mechanism is established, sharing relevant information through parallel storage tanks. Then, an adaptive dynamic adjustment algorithm for the storage tanks is designed to adapt to changes in the external operating environment. The established prediction model can simultaneously predict the energy consumption of two different main energy-consuming links, namely pumping energy consumption and aeration energy consumption, which alleviates the data waste caused by multi-rate phenomenon and utilizes useful information ignored by single-task learning. It can also adjust the size of the storage tank according to the dynamically changing operating conditions. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a flowchart of a power consumption prediction method based on a multi-task interpolation echo state network according to the present invention.

[0035] Figure 2 The figure shows the test results of the method for predicting pumping energy consumption tasks according to the present invention.

[0036] Figure 3 The figure shows the test results of the aeration task prediction method of the present invention.

[0037] Figure 4 The test error diagram shows the method for predicting pumping energy consumption tasks according to the present invention.

[0038] Figure 5 This is a test error diagram of the aeration energy consumption prediction method of the present invention;

[0039] Figure 6 This is a schematic diagram showing the detailed prediction method for aeration energy consumption in this invention. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0042] like Figure 1 As shown, this invention provides a power consumption prediction method based on a multi-task interpolation echo state network, comprising:

[0043] Step 100: Obtain the wastewater data to be tested;

[0044] Step 200: Using the synthesized time series, obtain the input data based on the wastewater data to be tested;

[0045] Step 300: Input the data to be input into the multi-task interpolation echo state network model to obtain the predicted power consumption value, wherein the predicted power consumption value is the pumping energy consumption value and the aeration energy consumption value, and the multi-task interpolation echo state network model includes an input layer, a storage tank and an output layer.

[0046] Furthermore, the wastewater data to be tested includes:

[0047] Effluent COD concentration, effluent total phosphorus concentration, influent pH, effluent pH, DO concentration.

[0048] Furthermore, it also includes:

[0049] Obtain the input data and predicted power consumption values;

[0050] The Laida criterion is used to detect outliers in the input data and the predicted power consumption. If outliers are found, they are replaced.

[0051] Specifically, such as Figure 6 As shown, variables strongly correlated with pumping energy consumption and aeration energy consumption from the data collected from the wastewater treatment plant were selected as inputs, including effluent COD concentration, effluent total phosphorus concentration, influent pH, effluent pH, and DO concentration. Pumping energy consumption and aeration energy consumption were selected as output variables. To eliminate the influence of factors such as sensor calibration and malfunctions, the Raida criterion was used to replace detected outliers with normal values ​​from the previous time step. The specific formula is as follows:

[0052] s o ={s∣s>(μ+3σ)∪s<(μ-3σ)} (1)

[0053] Where s o denoted as outliers, s represents the original data, and μ and σ represent the mean and standard deviation of the data, respectively. To remove the influence of different variables' dimensions on the prediction, the input and output variables are normalized to the [0,1] interval using the following formula.

[0054]

[0055] in s represents the normalized data. min and s max These represent the minimum and maximum values ​​of the data before normalization. The processed input data is denoted as u(t) = [u1(t), u2(t), u3(t), u4(t), u5(t)], and the output data is denoted as [y1(t), y2(t)]. Where u1(t), u2(t), u3(t), u4(t), and u5(t) represent the pretreated effluent COD concentration, effluent total phosphorus concentration, influent pH, effluent pH, and DO concentration, respectively. y1(t) and y2(t) represent the pretreated pumping energy consumption and aeration energy consumption, respectively.

[0056] Furthermore, acquiring the wastewater data to be tested includes:

[0057] Retrieve labeled and unlabeled data;

[0058] An interpolation method based on distance coefficients is used to obtain virtual interpolation points using unlabeled data;

[0059] The wastewater data to be tested is obtained based on the virtual interpolation points and the actual interpolation points corresponding to the labeled data.

[0060] Furthermore, the distance coefficient-based interpolation method, which uses unlabeled data to obtain virtual interpolation points, includes:

[0061] Determine the label data at the first time step and the label data at the second time step;

[0062] The average value of the unlabeled data between the first and second time points is obtained based on the label data at the first time point and the label data at the second time point.

[0063] The virtual interpolation point is determined based on the average value of the unlabeled data.

[0064] Specifically, this involves an interpolation method for multi-rate sampling data based on distance coefficients. Data collected from wastewater treatment plants is characterized by multi-rate sampling. This multi-rate phenomenon means that easily measurable variables, such as input variables like pH and DO concentration, are updated every τ minutes, while target variables, such as pumping energy consumption and aeration energy consumption, are updated every Kτ (K>1) minutes. This difference in update frequency results in some data being labeled and some being unlabeled, meaning only the values ​​of the input variables are present, without the corresponding values ​​of the output variables. The mathematical formula is as follows, where u... (i) This represents the input data, where i represents the type of sampling rate. This represents high-rate sampling input data, with sampling time rτ, meaning it can capture times τ, 2τ, ..., LKτ. This represents input data sampled at a low rate, with sampling time rKτ, meaning it can take values ​​of Kτ, 2Kτ, ..., LKτ. (rKτ) This represents the output data sampled at a low sampling rate. Its sampling time is rKτ, which can be Kτ, 2Kτ, ..., LKτ, where L represents the sample length and K represents how many times the high sampling rate is greater than the low sampling rate.

[0065]

[0066] y = {y (rKτ)} r＝1,2,…,L (4)

[0067] Therefore, this invention designs an interpolation method based on distance coefficients, which can utilize unlabeled data to generate new virtual interpolation points, thereby increasing the sample size of the model. The input data for the newly inserted auxiliary prediction points is the average of the unlabeled data between every two adjacent labeled data points, i.e., between times jKτ and (j+1)Kτ. time.

[0068] Furthermore, the step of using synthetic time series data to obtain input data based on the wastewater data to be tested includes:

[0069] The labeled time series is obtained based on the actual interpolation points corresponding to the labeled data;

[0070] A virtual time series is obtained based on the virtual interpolation points;

[0071] A synthesized time series is obtained based on the labeled time series and the virtual time series;

[0072] The input data is obtained based on the synthesized time series.

[0073] Specifically,

[0074] In order to solve Compute input data and historical labeled data Euclidean distance Calculate the weights based on the distance coefficient. The formula is as follows. Wherein, Indicates virtual input data with labeled time series European distance, This indicates the weight of the virtual output data.

[0075]

[0076] Virtual output data can be obtained as follows.

[0077]

[0078] The same method is used to generate it. At this point, a virtual sequence S2 is obtained, which can be represented as follows.

[0079]

[0080] The synthesized time series S, consisting of the labeled time series S1 and the virtual time series S2, can be represented as follows, where... The input data represents the synthesized time series S. and output data Its sampling time is From Kτ, …obtained LKτ.

[0081]

[0082] Furthermore, the network structure of the multi-task interpolation echo state network model is 5 - (k1N + k2N) - 2, where N is an initial positive integer, typically 10. j ≥1 and a positive integer, representing the number of sub-reservoir pools owned by each ESN module, k j N represents the size of the storage tank for the pumping energy consumption of Task 1 and the aeration energy consumption of Task 2, respectively, where j = 1, 2.

[0083] The multi-task interpolation echo state network consists of three parts: an input layer, a reservoir, and an output layer. The input layer has five input variables: effluent COD concentration, effluent total phosphorus concentration, influent pH, effluent pH, and DO concentration. The reservoir is formed by two parallel ESN modules, each containing several sub-reservoirs with identical nodes, possessing multi-task learning capabilities to capture and memorize input sequence information. The output layer provides predicted pumping and aeration energy consumption, using only real-time values ​​to evaluate prediction accuracy. Therefore, the neural network model has five input variables and two output variables, resulting in a network structure of 5 - (k1N + k2N) - 2, where N is an initial positive integer, typically 10. j ≥1 (j=1,2) and a positive integer, representing the number of sub-reservoir pools owned by each ESN module, k j N(j=1,2) represent the size of the storage tank for the pumping energy consumption of Task 1 and the aeration energy consumption of Task 2, respectively.

[0084] For the d-th ESN module, assume Sub-reservoir pools, each with N nodes. For the d-th node... k (k = 1, 2, ..., k) d Sub-reservoir, with input weight matrix as follows: The sub-reservoir weight matrix is Based on singular value decomposition, using a predefined diagonal matrix and two randomly constructed orthogonal matrices and generate and These are the output weights for the pumping energy consumption of Task 1 and the aeration energy consumption of Task 2, respectively. At time step t, the internal state matrix Harmony reserve pool output It can be calculated by the following formula

[0085]

[0086] Where f(·) is the activation function, typically the hyperbolic tangent function tanh(), and u(t) is the input signal. From the perspective of the d-th ESN module, its input weights are... Reserve Pool Weight It can be represented as Reserve pool status X d (t+1) and the corresponding output O d The (t+1) update is as follows. Among them, It is the dth module. kThe output weights of each sub-reservoir pool It is the dth module. k The status of individual reserve pools.

[0087]

[0088] Let X(t) = [X1(t); X2(t)], and let the network output O of task p (p = 1, 2) be O. p (t) is represented as

[0089]

[0090] in It is the output weight of the reservoir of output node p. Formula (16) is restated as follows:

[0091]

[0092] The solution is obtained using the gradient descent method, as shown in equation (18). p (t) represents the training modeling error, E p (t) is the loss function. It is the corresponding supervision signal matrix, and y1 and y2 are the teacher signals during the training phase, namely the pumping energy consumption and aeration energy consumption to be predicted.

[0093]

[0094] e p (t)=Y p (t)-O p (t) (19)

[0095]

[0096] Where η is the learning rate. L1 represents the corresponding gradient. L1 is the length of the training samples.

[0097] Formulas 12-21 together constitute the multi-task interpolation echo state network model.

[0098] Furthermore, the network is trained:

[0099] Initialize two ESN modules for the two tasks of the network. Each module has a sub-reservoir with N=10 nodes. Randomly generate the input weight matrix and internal connection matrix for each sub-reservoir.

[0100] Input a training data sample In ESN modules 1 and 2, data processing is performed using a multi-rate data interpolation method to obtain... and Collection status renew Calculate the training modeling error e p (t).

[0101] Calculate the comprehensive modeling error E(t) for each time step, for the d-th module. k Each sub-pool is used to assess its learning ability for each task, using contribution. As follows. Among them, This represents the dth module of the p-th task. k The output weights of each sub-reservoir pool Represents the d-th module k The status of individual reserve pools.

[0102]

[0103] To measure the predictive ability of the d-th module for task p, a cumulative contribution is introduced. as follows.

[0104]

[0105] To achieve a compact network structure, the design principles of a dynamic structure adjustment strategy are presented.

[0106] Growth mechanism: To compensate for large modeling errors, new sub-reservoir pools are added. The network growth condition is expressed by the following equation.

[0107] E(t)>E a and k d N≤N max (25)

[0108] Where E a and N max E is a predefined positive value. a N represents the maximum modeling error in practical applications, determined based on the specific task. max This represents the maximum size of a single module reservoir, typically set to 200. Then, the module α with the smallest cumulative contribution from Task 1 and Task 2, and the sub-reservoir α with the largest contribution are selected. h .

[0109]

[0110] Next, the new sub-reservoir is initialized and inserted into module α. Input weights are then assigned to the new neurons. Internal weight ΔW new (t) and output weights As follows, at this time k d =k d +1.

[0111]

[0112] Pruning Mechanism: To achieve a concise network, a node pruning algorithm is designed by eliminating sub-pools with smaller contributions. First, the sub-pools are sorted from highest to lowest contribution. Then, the sub-pool with the smallest contribution to both Task 1 and Task 2 is selected and named β. h In the ESN module β, the sub-reservoir that contributes the second smallest amount is denoted as β. h' If the following conditions are met,

[0113]

[0114] Where γ>0 is the deletion threshold, set to γ=0.05. Sub-reservoir β h and β h' The weights are adjusted as follows.

[0115]

[0116] Stability strategy: The network structure remains unchanged under a fixed reserve pool once the following conditions are met.

[0117]

[0118] Repeat steps to update Each update The number of training iterations is i = i + 1, until the maximum number of training iterations i = i is reached. max Training stops when the value reaches L1.

[0119] After training, the output of the neural network is inversely normalized to obtain predicted estimates of pumping energy consumption and aeration energy consumption using the following formula:

[0120]

[0121] Going further, let's test the neural network:

[0122] After removing outliers and normalizing the test sample data according to formulas (1)-(2), the interpolated synthetic sequence is obtained according to formulas (5)-(11) and used as the input of the trained multi-task interpolation echo state network. The output of the multi-task interpolation echo state network is inversely normalized according to formula (31) to obtain the estimated measured values ​​of pumping energy consumption and aeration energy consumption.

[0123] More specifically, using the 2023 water quality and energy consumption analysis data from a wastewater treatment plant in Beijing, a total of 1000 data samples were obtained. Of these, 500 were labeled samples and 500 were unlabeled samples. Labeled samples are those where the input data is updated with corresponding output data, while unlabeled samples only contain input data without corresponding output data. The first 700 data samples were selected as training samples, and the last 300 data samples were selected as test samples. The first 700 data samples included 350 labeled samples, and the last 300 data samples included 150 labeled samples. Only the labeled samples were used to calculate the test error. The main steps are as follows:

[0124] Step 1: Selection of input variables and data preprocessing for the soft measurement model.

[0125] The effluent COD concentration, effluent total phosphorus concentration, influent pH, effluent pH, and DO concentration were selected as input variables, and pumping energy consumption and aeration energy consumption were selected as output variables. Outliers were removed using formula (1), and the input and output variables were normalized to the [0,1] interval using formula (2).

[0126] Step 2: Initialize the multi-task interpolation echo state network structure and parameters.

[0127] First, the multi-task interpolation echo state network structure is determined to be 5 - (k1N + k2N) - 2; second, the input weights of each sub-reservoir are initialized. and internal weights The value is controlled between [-0.5, 0.5], and the maximum singular value is... Finally, the network updates its state according to formulas (12) and (14), and calculates the output according to formula (16).

[0128] Step 3: Train the neural network.

[0129] Step 3.1: Initialize the two ESN modules for the two tasks of the network. Each module has a sub-reservoir and N=10 nodes. Randomly generate the input weight matrix and internal connection matrix for each sub-reservoir.

[0130] Step 3.2: Input a training data sample In ESN modules 1 and 2, data processing is performed using a multi-rate data interpolation method to obtain... and The state of each sub-reservoir is collected using formulas (12) and (14). and the state X of each module d (t+1), updated via formula (18) The training modeling error e is calculated using formula (19). p (t).

[0131] Step 3.3: Calculate the comprehensive modeling error E(t) for each time step using formula (22), and calculate the contribution of each sub-reservoir and the cumulative contribution of each module using formulas (23) and (24). If E(t) satisfies formula (25), then a new sub-reservoir is added under the specified module according to formulas (26) and (27). If formula (28) is satisfied, then the sub-reservoir β is adjusted according to formula (29). h and β h' The weight.

[0132] Step 3.4: Repeat steps 3.2-3.3 to update. Each update The number of training iterations is i = i + 1, until the maximum number of training iterations i = i is reached. max Training stops when the value reaches L1.

[0133] Step 3.5: After training, the output of the neural network is inversely normalized to the predicted values ​​of pumping energy consumption and aeration energy consumption using formula (31).

[0134] Step 4: Test the neural network.

[0135] The pumping energy consumption and aeration energy consumption were tested using the trained network. The test results for pumping energy consumption are as follows: Figure 2 As shown, Figure 2 X-axis: Time step Figure 2 Y-axis: Pumping energy consumption value, unit is kWh; solid line is the expected pumping energy consumption value, dashed line is the tested pumping energy consumption output value; the test results of aeration energy consumption are as follows: Figure 3 As shown, Figure 3 X-axis: Time step Figure 3 Y-axis: Aeration energy consumption value, unit is kWh; solid line is the expected aeration energy consumption value, dashed line is the tested aeration energy consumption output value; the error between the actual output and the tested output of pumping energy consumption is as follows: Figure 4 As shown, Figure 4 X-axis: Time step Figure 4 Y-axis: Pumping energy consumption error, unit is kWh; the error between actual and test output of aeration energy consumption is as follows: Figure 5 As shown, Figure 5 X-axis: Time step Figure 5 Y-axis: Aeration energy consumption error, in kWh; Experimental results demonstrate the effectiveness of the prediction model for pumping energy consumption and aeration energy consumption based on a multi-task interpolation echo state network.

[0136] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0137] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A power consumption prediction method based on a multi-task interpolation echo state network, characterized in that, include: Acquire wastewater data; Using a synthetic time series, the input data is obtained from the wastewater data to be tested; The input data is input into the multi-task interpolation echo state network model to obtain the predicted power consumption value, wherein the predicted power consumption value is the pumping energy consumption value and the aeration energy consumption value, and the multi-task interpolation echo state network model includes an input layer, a storage tank and an output layer. The network structure of the multi-task interpolation echo state network model is as follows: Where N is an initial positive integer, usually 10. And it is a positive integer, representing each The number of sub-reservoir pools that the module possesses. The sizes of the storage tanks represent the energy consumption for pumping in Task 1 and the energy consumption for aeration in Task 2, respectively. ; The expression for the model of the multi-task interpolation echo state network is: ； in, The original data, For the normalized data, and These are the minimum and maximum values ​​of the input data before normalization, respectively.

2. The power consumption prediction method based on a multi-task interpolation echo state network according to claim 1, characterized in that, The wastewater data to be tested includes: Effluent COD concentration, effluent total phosphorus concentration, influent pH, effluent pH, DO concentration.

3. The power consumption prediction method based on a multi-task interpolation echo state network according to claim 1, characterized in that, The acquisition of wastewater data to be tested includes: Retrieve labeled and unlabeled data; An interpolation method based on distance coefficients is used to obtain virtual interpolation points using unlabeled data; The wastewater data to be tested is obtained based on the virtual interpolation points and the actual interpolation points corresponding to the labeled data.

4. The power consumption prediction method based on a multi-task interpolation echo state network according to claim 3, characterized in that, The distance coefficient-based interpolation method, which uses unlabeled data to obtain virtual interpolation points, includes: Determine the label data at the first time step and the label data at the second time step; The average value of the unlabeled data between the first and second time points is obtained based on the label data at the first time point and the label data at the second time point. The virtual interpolation point is determined based on the average value of the unlabeled data.

5. The power consumption prediction method based on a multi-task interpolation echo state network according to claim 3, characterized in that, The process of obtaining input data from the wastewater data using synthesized time series data includes: The labeled time series is obtained based on the actual interpolation points corresponding to the labeled data; A virtual time series is obtained based on the virtual interpolation points; A synthesized time series is obtained based on the labeled time series and the virtual time series; The input data is obtained based on the synthesized time series.

6. The power consumption prediction method based on a multi-task interpolation echo state network according to claim 1, characterized in that, Also includes: Obtain the input data and predicted power consumption values; The Laida criterion is used to detect outliers in the input data and the predicted power consumption. If outliers are found, they are replaced.

Images