A method for measuring total nitrogen in effluent from a sewage treatment plant

Abstract

The present invention discloses a method for measuring total nitrogen in effluent from a sewage treatment plant, comprising S1: constructing a multi-temporal and spatial scale of the sewage treatment plant according to the process flow of the sewage treatment plant; the multi-temporal and spatial scale of the sewage treatment plant includes a plant-level temporal and spatial scale and a process-level temporal and spatial scale; S2: constructing a sewage treatment plant operation condition identification model based on the plant-level temporal and spatial scale, and constructing a biological pool operation condition identification model based on the process-level temporal and spatial scale; S3: selecting input variables according to the correlation relationship of the process flow of the sewage treatment plant, analyzing them through an LSTM model, and obtaining a final output; S4: constructing a multi-temporal and spatial scale condition model according to the sewage treatment plant operation condition identification model and the biological pool operation condition identification model, thereby outputting the measured value of the total nitrogen at the sewage treatment plant outlet.

Description

Technical Field

[0001] The invention relates to the technical field of environmental protection, and in particular to a method for measuring total nitrogen in effluent from a sewage plant. Background Art

[0002] Water is the basic material condition for human survival and social development, and is a limited, irreplaceable and precious resource. As an important part of water resource protection, sewage treatment uses physical, chemical and biological methods to remove organic matter, ammonia nitrogen and other pollutants from sewage generated by urban life and industrial production, so as to achieve the purpose of sewage purification and recycling.

[0003] The sewage treatment process is a complex dynamic biochemical process with strong external interference, strong time-varying, strong coupling and nonlinearity. Its ultimate goal is to make the purified effluent water quality meet the specified indicators. The activated sludge method is the most widely used sewage treatment process in sewage treatment. It uses the biological characteristics of microbial flora to decompose and oxidize the soluble organic matter in sewage through biochemical reactions such as nitrification and denitrification, so that the sewage is purified and meets the national emission standards. The sewage effluent water quality index is an important indicator to measure whether the sewage treatment process is normal or not, and it is also an indicator of the specific state changes within the reaction process. Common effluent water quality includes biochemical oxygen demand (BOD), chemical oxygen demand (COD) and total suspended solids (TSS). However, due to the influence of influent flow, microbial population, dissolved oxygen concentration, pH value, etc., it is difficult to achieve real-time online detection of effluent water quality indicators, and offline laboratory measurement has a large lag, which is easy to cause secondary pollution.

[0004] In order to achieve real-time, comprehensive and accurate detection of the internal operating status of the biochemical reactor and the secondary sedimentation tank, it is necessary to establish an accurate and reliable soft measurement model that can reflect the trend of the sewage effluent water quality indicators in the present and future. Measurement technology is a common new technology used in the field of process control to solve the problem of difficult-to-measure parameter detection. By establishing easy-to-measure auxiliary variables, the estimation of difficult-to-measure variables can be achieved, which can effectively solve the measurement problem of difficult-to-measure variables (total nitrogen in effluent).

[0005] There is currently no soft measurement applied in sewage treatment plants. The main reason is that sewage treatment plants belong to biological treatment processes, and the measurement cycle of water quality parameters is long, making it difficult to achieve rapid measurement; secondly, sewage treatment plants have many operating conditions, making it difficult to clearly determine the operating conditions. Summary of the invention

[0006] In view of the technical problem of low measurement accuracy of total nitrogen in sewage treatment plant effluent in the prior art, the present invention proposes a method for measuring total nitrogen in sewage treatment plant effluent, which effectively solves the detection of difficult-to-measure parameters, improves the perception of water quality variables at the sewage treatment plant outlet, and provides theoretical methodological support for the soft measurement method of total nitrogen at the sewage treatment plant outlet driven by multi-temporal and spatial scale data.

[0007] In order to achieve the above object, the present invention provides the following technical solutions:

[0008] A method for measuring total nitrogen in effluent from a sewage treatment plant, comprising the following steps:

[0009] S1: According to the process flow of the sewage treatment plant, the multi-temporal and spatial scale of the sewage treatment plant is constructed; the multi-temporal and spatial scale of the sewage treatment plant includes the plant-level temporal and spatial scale and the process-level temporal and spatial scale;

[0010] S2: Construct a sewage treatment plant operation condition identification model based on the plant-level spatiotemporal scale, and construct a biological pool operation condition identification model based on the process-level spatiotemporal scale;

[0011] S3: Select input variables according to the correlation of the sewage treatment plant process, analyze them through the LSTM model, and obtain the final output;

[0012] S4: Construct a multi-temporal and spatial scale operating condition model and the final output based on the sewage treatment plant operating condition identification model and the biological pool operating condition identification model, thereby outputting the measured value of total nitrogen at the sewage treatment plant outlet.

[0013] Preferably, in S1, the process flow of sewage treatment in the sewage treatment plant is as follows: sewage enters the sewage treatment plant through the sewage treatment plant entrance, is treated in various processes, and is discharged at the sewage treatment plant outlet after various indicators meet the discharge standards.

[0014] Preferably, in S1, the plant-level spatiotemporal scale includes the state change of sewage from the inlet to the outlet of the sewage treatment plant, the power consumption of the blower, and the drug consumption; the process-level spatiotemporal scale includes the state change of sewage from the inlet to the outlet of each process.

[0015] Preferably, in S2, the sewage treatment plant operation condition identification model L is:

[0016] L=K1*power consumption / influent flow rate+K2*drug consumption / influent flow rate+K3*effluent total nitrogen / influent total nitrogen+K4*effluent ammonia nitrogen / influent ammonia nitrogen+K5*effluent COD / influent COD;

[0017] Among them, K1 represents the weight coefficient of electricity consumption per unit flow rate; K2 represents the weight coefficient of drug consumption per unit flow rate; K3 represents the weight coefficient of the ratio of total nitrogen in inlet and outlet water; K4 represents the weight coefficient of the ratio of ammonia nitrogen in inlet and outlet water; K5 represents the weight coefficient of the ratio of COD inlet and outlet water.

[0018] Preferably, in S2, the biological pool operation condition identification model G is:

[0019] G = M1*DO / set value of dissolved oxygen concentration + M2*MLSS / reference value of sludge concentration + M3*ammonia nitrogen / 10+M4*XD / 15+M5*air flow / 2000;

[0020] Among them, M1 represents the weight coefficient of the ratio of the actual value of dissolved oxygen concentration in the biological pool to the set value; M2 represents the weight coefficient of the ratio of the actual value of sludge concentration in the biological pool to the benchmark value; M3 represents the weight coefficient of the actual value of ammonia nitrogen at the outlet of the biological pool; M4 represents the weight coefficient of nitrate nitrogen XD at the outlet of the biological pool; M5 represents the weight coefficient of the air flow rate at the inlet of the biological pool.

[0021] Preferably, the S3 includes:

[0022] S3-1: Select input variables based on the sewage treatment plant process flow;

[0023] S3-2: Use LSTM to analyze the input variables and get the final output.

[0024] Preferably, in S3-1, the input variables include nitrate nitrogen in the biological pool effluent, ammonia nitrogen in the biological pool effluent, nitrate nitrogen in the biological pool inlet, inlet flow rate, biological pool sludge concentration MLSS, and blower air flow rate.

[0025] Preferably, the S3-2 includes:

[0026] S3-2-1: First, pass the forget gate to convert the hidden state h of the previous moment t-1 and the current input x t The first sigmoid layer outputs the degree of information to be forgotten f t :

[0027] f t =σ 1 (W f ·[h t-1 ,x t ]+b f ) (1)

[0028] In formula (1), f t Indicates the degree of forgotten information; σ 1 W represents the forget gate; f represents the weight matrix of the forget gate; h t-1 represents the hidden state at time t-1; x t represents the input at time t, i.e., the input variable in S3-1, b f Represents the bias term of the forget gate;

[0029] S3-2-2: Then the second sigmoid layer and tanh layer in the input gate output the updated data i respectively t and candidate values

[0030]

[0031] In formula (2), i t Represents the updated data at time t; σ 2 represents the input gate; W i represents the weight matrix of the input gate, W c represents the memory weight matrix, b i represents the bias term of the input gate, b c represents the bias term of memory; tanh represents the tanh layer; represents a candidate value;

[0032] S3-2-3: Combined update data i t and candidate values Determine the updated input data C t :

[0033]

[0034] In formula (3), C t Indicates the unit state of the current input, C t-1 Indicates the unit state input at the last moment;

[0035] S3-2-4: Update the input data C t Input and output gate, get the intermediate value o through the third sigmoid layer t , and then the final output h is obtained through the tanh function t :

[0036] o t =σ 3 (W o ·[h t-1 ,x t ]+b o ), h t =o t *tanh(C t ) (4)

[0037] In formula (4), o t represents the middle value; σ 3 represents the third sigmoid layer; W o represents the weight matrix of the output gate, b o represents the bias term of the output gate; tanh represents the tanh function; h t Represents the final output.

[0038] Preferably, the S4 includes:

[0039] S4-1: Construct a multi-time and space scale operating model Z = (L, G);

[0040] S4-2: Based on the operating condition combination of Z, output the measured value Y of total nitrogen at the sewage treatment plant outlet.

[0041] Preferably, in S4-2, the measured value Y of total nitrogen at the sewage treatment plant outlet is:

[0042] Y=h t *In (5)

[0043] In formula (5), Y represents the measured value of total nitrogen at the outlet of the sewage treatment plant; h t Represents the final output of the LSTM model; In represents the coefficient of the nth operating condition.

[0044] In summary, due to the adoption of the above technical solution, compared with the prior art, the present invention has at least the following beneficial effects:

[0045] The present invention is innovative in that it identifies the operating conditions through spatiotemporal scales, and selects easy-to-measure auxiliary variables to detect difficult-to-measure coefficients, thereby effectively solving the detection of difficult-to-measure parameters, improving the perception of water quality variables at the sewage treatment plant outlet, and providing theoretical methodological support for the soft measurement method of total nitrogen at the sewage treatment plant outlet driven by multi-spatiotemporal scale data. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 Schematic diagram of a method for measuring total nitrogen in effluent from a sewage treatment plant according to an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0047] The present invention is further described in detail below in conjunction with the examples and specific implementation methods. However, this should not be understood as the scope of the above subject matter of the present invention being limited to the following examples, and all technologies realized based on the content of the present invention belong to the scope of the present invention.

[0048] In the description of the present invention, it is necessary to understand that the terms "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0049] like Figure 1As shown, the present invention provides a method for measuring total nitrogen in effluent from a sewage treatment plant, comprising the following steps:

[0050] S1: Construct multiple spatiotemporal scales of the sewage treatment plant based on the process flow of the sewage treatment plant.

[0051] In this embodiment, the process flow of sewage treatment in the sewage treatment plant is as follows: sewage enters the sewage treatment plant through the sewage treatment plant entrance, is treated in various processes, and is discharged at the sewage treatment plant outlet after various indicators meet the discharge standards.

[0052] Therefore, according to the technological process of the sewage treatment plant, the sewage treatment plant is divided into plant-level time and space scale and process-level time and space scale.

[0053] In this embodiment, the plant-level spatiotemporal scale includes the state changes of sewage from the inlet to the outlet of the sewage treatment plant, the power consumption and drug consumption of the blower, etc.; the process-level spatiotemporal scale includes the state changes of sewage from the inlet to the outlet of each process.

[0054] S2: Construct a sewage treatment plant operating condition identification model based on the plant-level spatiotemporal scale; construct a biological pool operating condition identification model based on the process-level spatiotemporal scale.

[0055] S2-1: Construct a sewage treatment plant operating condition identification model based on the parameters in the plant-level time and space scale (including inlet water flow, inlet ammonia nitrogen, inlet total nitrogen, inlet chemical oxygen demand COD, outlet water flow, effluent total nitrogen, effluent ammonia nitrogen, effluent chemical oxygen demand COD, and the power consumption and drug consumption of the whole plant during this period. The above parameters can be obtained from sensors and are known parameters).

[0056] In this embodiment, the sewage treatment plant operation condition identification model L is:

[0057] L=K1*power consumption / influent flow rate+K2*drug consumption / influent flow rate+K3*effluent total nitrogen / influent total nitrogen+K4*effluent ammonia nitrogen / influent ammonia nitrogen+K5*effluent COD / influent COD;

[0058] Among them, K1 represents the weight coefficient of electricity consumption per unit flow rate; K2 represents the weight coefficient of drug consumption per unit flow rate; K3 represents the weight coefficient of the ratio of total nitrogen in inlet and outlet water; K4 represents the weight coefficient of the ratio of ammonia nitrogen in inlet and outlet water; K5 represents the weight coefficient of the ratio of COD inlet and outlet water.

[0059] In this embodiment, according to the range of the sewage treatment plant operating condition identification model L (by calculation, the calculation result is between 0-100, and the operating conditions are divided according to the normal distribution), the plant-level operating conditions are divided into five types: good, better, normal, general, and abnormal.

[0060] S2-2: Construct a biological pool operation condition identification model based on the process-level spatiotemporal scale.

[0061] In this embodiment, the process-level spatiotemporal scales include dissolved oxygen concentration DO in the biological pond, sludge concentration MLSS, nitrate nitrogen XD, ammonia nitrogen NH, air flow, etc., and these data are known.

[0062] Among them, the biological pool operation condition identification model G is:

[0063] G = M1*DO / set value of dissolved oxygen concentration + M2*MLSS / reference value of sludge concentration + M3*ammonia nitrogen / 10+M4*XD / 15+M5*air flow / 2000;

[0064] Among them, M1 represents the weight coefficient of the ratio of the actual value of dissolved oxygen concentration in the biological pool to the set value; M2 represents the weight coefficient of the ratio of the actual value of sludge concentration in the biological pool to the benchmark value; M3 represents the weight coefficient of the actual value of ammonia nitrogen at the outlet of the biological pool; M4 represents the weight coefficient of nitrate nitrogen XD at the outlet of the biological pool; M5 represents the weight coefficient of the air flow rate at the inlet of the biological pool.

[0065] In this embodiment, according to the range of the process-level working condition identification model G (by calculating so that the calculation result is between 0-100, and the working conditions are divided according to the normal distribution), the process-level working conditions are divided into five types: good, better, normal, general, and abnormal.

[0066] S3: Select input variables according to the correlation of the sewage plant process flow, analyze them through the LSTM model, and get the final output.

[0067] S3-1: Select the input variables of the sewage treatment plant process.

[0068] In this embodiment, the input variables are selected from an optimal subset of the original feature space (all parameters of the sewage treatment plant), and the subset retains the entire or most of the information of the original feature set, reducing the correlation between variables, thereby reducing the dimension of the new feature space.

[0069] In this embodiment, the selected input variables are highly correlated with the total nitrogen in the effluent, including nitrate nitrogen in the effluent of the biological pool, ammonia nitrogen in the effluent of the biological pool, nitrate nitrogen at the inlet of the biological pool, inlet flow rate, MLSS concentration of sludge in the biological pool, air flow rate of the blower, etc.

[0070] In this embodiment, the min-max method is used to normalize the selected variables:

[0071]

[0072] Among them, x' represents the input variable after normalization; x represents the input variable before normalization; max(x) represents the maximum value of the input variable before normalization; min(x) represents the minimum value of the input variable before normalization.

[0073] S3-2: Use LSTM (Long Short Term Memory Network) to analyze the input variables and alleviate the problem of gradient disappearance by capturing the long-term dependencies in the input variables, thereby obtaining the final output.

[0074] S3-2-1: A single LSTM neural network unit first passes through the forget gate to pass the hidden state h of the previous moment t-1 and the current input x t The first sigmoid layer outputs the degree of information to be forgotten f t , if f t 1 means that the long-term memory state of the previous moment is completely retained. t A value of 0 means that the long-term memory state of the previous moment is completely discarded.

[0075] f t =σ 1 (W f ·[h t-1 ,x t ]+b f ) (1)

[0076] In formula (1), f t Indicates the degree of forgotten information; σ 1 W represents the forget gate (the first sigmoid layer); f represents the weight matrix of the forget gate; h t-1 represents the hidden state at time t-1; x t represents the input at time t, i.e., the input variable in S3-1; b f Represents the bias term of the forget gate.

[0077] S3-2-2: Then the second sigmoid layer and tanh layer in the input gate output the updated data i respectively t and candidate values This is used to jointly decide what information to store in the cell state.

[0078]

[0079] In formula (2), i t Represents the updated data at time t; σ 2 represents the input gate (the second sigmoid layer); W i represents the weight matrix of the input gate, W c represents the memory weight matrix, b i represents the bias term of the input gate, b c represents the bias term of memory; tanh represents the tanh layer; represents a candidate value;

[0080] S3-2-3: According to the degree of forgotten information f t Determine the forgotten information (i.e. the output state of the unit at the previous moment, implemented according to the weight matrix of the forget gate); then combine the updated data i t and candidate values Determine the updated input data C t .

[0081] In this embodiment, the old state C t-1 Multiply by the output of the forget gate f t , decide to forget the information. Then combine the updated data i output by the input gate t and candidate values Together they determine the new input, thereby calculating the new cell state C t ; some parts of the old state will be forgotten and new candidate values ​​will be added.

[0082]

[0083] In formula (3), C t Represents updated input data; C t-1 Indicates the unit status input at the last moment.

[0084] S3-2-4: Update the input data C t Input and output gate, get the intermediate value o through the third sigmoid layer t , and then the final output h is obtained through the tanh function t , and use this value as the input of the next time step LSTM for network prediction.

[0085] o t =σ 3 (W o ·[h t-1 ,x t ]+b o ), h t =o t *tanh(C t ) (4)

[0086] In formula (4), o t represents the middle value; σ 3 represents the third sigmoid layer; W o represents the weight matrix of the output gate, b o represents the bias term of the output gate; tanh represents the tanh function; h t Represents the final output.

[0087] S4: Construct a multi-temporal and spatial scale operating condition model Z based on the sewage treatment plant operating condition identification model L and the biological pool operating condition identification model G, thereby outputting the measured value of total nitrogen at the sewage treatment plant outlet.

[0088] S4-1: Construct a multi-time and space scale operating condition model Z = (L, G). According to the operating conditions of L and G, 20 operating condition combinations can be formed. Each operating condition is set with a coefficient In, where n is the operating condition combination number and takes a value of 1-20.

[0089] In this embodiment, for example, the following working conditions are included:

[0090] Operating condition 1: The whole plant working condition is good and the biological pool working condition is good; Operating condition 2: The whole plant working condition is good and the biological pool working condition is relatively good; Operating condition 3: The whole plant working condition is good and the biological pool working condition is normal; Operating condition 4: The whole plant working condition is good and the biological pool working condition is general; Operating condition 5: The whole plant working condition is good and the biological pool working condition is abnormal; Operating condition 6: The whole plant working condition is relatively good and the biological pool working condition is good; Operating condition 7: The whole plant working condition is relatively good and the biological pool working condition is relatively good; Operating condition 8: The whole plant working condition is relatively good and the biological pool working condition is normal; Operating condition 9: The whole plant working condition is relatively good and the biological pool working condition is general; Operating condition 10: The whole plant working condition is relatively good and the biological pool working condition is abnormal, etc.

[0091] S4-2: According to the operating condition combination of Z, the measured value Y of total nitrogen at the outlet of the sewage treatment plant is output:

[0092] Y=h t *In (5)

[0093] In formula (5), Y represents the measured value of total nitrogen at the outlet of the sewage treatment plant; h t Represents the final output of the LSTM model; In represents the coefficient of the nth operating condition.

[0094] Those skilled in the art will appreciate that the above-mentioned embodiments are specific examples for implementing the present invention, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present invention.

Claims

1. A method for measuring total nitrogen in effluent from a sewage treatment plant, characterized in that , specifically including the following steps: S1: According to the process flow of the sewage treatment plant, the multi-temporal and spatial scale of the sewage treatment plant is constructed; the multi-temporal and spatial scale of the sewage treatment plant includes the plant-level temporal and spatial scale and the process-level temporal and spatial scale; S2: Construct a sewage treatment plant operation condition identification model based on the plant-level spatiotemporal scale, and construct a biological pool operation condition identification model based on the process-level spatiotemporal scale; S3: Select input variables according to the correlation of the sewage treatment plant process, analyze them through the LSTM model, and obtain the final output; S4: Construct a multi-temporal and spatial scale operating condition model and the final output based on the sewage treatment plant operating condition identification model and the biological pool operating condition identification model, thereby outputting the measured value of total nitrogen at the sewage treatment plant outlet.

2. A method for measuring total nitrogen in sewage treatment plant effluent according to claim 1, characterized in that ,In S1, the process flow of sewage treatment in the sewage plant is as follows: the sewage enters the sewage plant through the sewage plant entrance, is treated in various processes, and after all indicators meet the discharge standards, it is discharged at the sewage plant outlet.

3. A method for measuring total nitrogen in sewage treatment plant effluent according to claim 1, characterized in that In S1, the plant-level spatiotemporal scale includes the state change of sewage from the sewage plant entrance to the sewage plant exit, the power consumption of the blower, and the drug consumption; the process-level spatiotemporal scale includes the state change of sewage from the entrance to the exit of each process.

4. A method for measuring total nitrogen in sewage treatment plant effluent according to claim 1, characterized in that ,In S2, the sewage treatment plant operation condition identification model L is: L=K1*power consumption / influent flow rate+K2*drug consumption / influent flow rate+K3*effluent total nitrogen / influent total nitrogen+K4*effluent ammonia nitrogen / influent ammonia nitrogen+K5*effluent COD / influent COD; Among them, K1 represents the weight coefficient of electricity consumption per unit flow rate; K2 represents the weight coefficient of drug consumption per unit flow rate; K3 represents the weight coefficient of the ratio of total nitrogen in inlet and outlet water; K4 represents the weight coefficient of the ratio of ammonia nitrogen in inlet and outlet water; K5 represents the weight coefficient of the ratio of COD inlet and outlet water.

5. A method for measuring total nitrogen in sewage treatment plant effluent according to claim 1, characterized in that In S2, the biological pool operation condition identification model G is: G = M1*DO / set value of dissolved oxygen concentration + M2*MLSS / reference value of sludge concentration + M3*ammonia nitrogen / 10+M4*XD / 15+M5*air flow / 2000; Among them, M1 represents the weight coefficient of the ratio of the actual value of dissolved oxygen concentration in the biological pool to the set value; M2 represents the weight coefficient of the ratio of the actual value of sludge concentration in the biological pool to the benchmark value; M3 represents the weight coefficient of the actual value of ammonia nitrogen at the outlet of the biological pool; M4 represents the weight coefficient of nitrate nitrogen XD at the outlet of the biological pool; M5 represents the weight coefficient of the air flow rate at the inlet of the biological pool.

6. A method for measuring total nitrogen in sewage treatment plant effluent according to claim 1, characterized in that , said S3 comprises: S3-1: Select input variables based on the sewage treatment plant process flow; S3-2: Use LSTM to analyze the input variables and get the final output.

7. A method for measuring total nitrogen in sewage treatment plant effluent as claimed in claim 6, characterized in that In S3-1, the input variables include nitrate nitrogen in the biological pool effluent, ammonia nitrogen in the biological pool effluent, nitrate nitrogen in the biological pool inlet, inlet flow, biological pool sludge concentration MLSS, and blower air flow.

8. A method for measuring total nitrogen in sewage treatment plant effluent as claimed in claim 6, characterized in that , the S3-2 includes: S3-2-1: First, pass the forget gate to convert the hidden state h of the previous moment t-1 and the current input x t The first sigmoid layer outputs the degree of information to be forgotten f t : f t =σ1(W f ·[h t-1 ,x t ]+b f ) (1) In formula (1), f t Indicates the degree of forgetting information; σ1 represents the forget gate; W f represents the weight matrix of the forget gate; h t-1 represents the hidden state at time t-1; x t represents the input at time t, i.e., the input variable in S3-1, b f Represents the bias term of the forget gate; S3-2-2: Then the second sigmoid layer and tanh layer in the input gate output the updated data i respectively t and candidate values In formula (2), i t represents the updated data at time t; σ2 represents the input gate; W i represents the weight matrix of the input gate, W c represents the memory weight matrix, b i represents the bias term of the input gate, b c represents the bias term of memory; tanh represents the tanh layer; represents a candidate value; S3-2-3: Combined update data i t and candidate values Determine the updated input data C t : In formula (3), C t Indicates the unit state of the current input, C t-1 Indicates the unit state input at the last moment; S3-2-4: Update the input data C t Input and output gate, get the intermediate value o through the third sigmoid layer t , and then the final output h is obtained through the tanh function t : o t =σ3(W o ·[h t-1 ,x t ]+b o ),h t =o t *tanh(C t ) (4) In formula (4), o t represents the middle value; σ3 represents the third sigmoid layer; W o represents the weight matrix of the output gate, b o represents the bias term of the output gate; tanh represents the tanh function; h t Represents the final output.

9. A method for measuring total nitrogen in sewage treatment plant effluent according to claim 1, characterized in that , the S4 includes: S4-1: Construct a multi-time and space scale operating model Z = (L, G); S4-2: Based on the operating condition combination of Z, output the measured value Y of total nitrogen at the sewage treatment plant outlet.

10. A method for measuring total nitrogen in effluent from a sewage treatment plant as claimed in claim 9, characterized in that In S4-2, the measured value Y of total nitrogen at the outlet of the sewage treatment plant is: Y=h t *In (5) In formula (5), Y represents the measured value of total nitrogen at the outlet of the sewage treatment plant; h t Represents the final output of the LSTM model; In represents the coefficient of the nth operating condition.

Images