JP4141815B2 - Plant operation optimization controller - Google Patents

Description

なお、ａ_１，ａ_２・・・は自己回帰のパラメータであり、予め与えることも可能であり、或いは、実時間で逐次最小２乗推定（カルマンフィルタ）することも可能である。
【００４９】
このように、上記（２）式で当日の予測値が得られ、また、その曜日Ｗの１日分のパターンが得られるため、各時間帯の比で積分すると、当日の１時間毎２４時間分の予測値は次のように得られる。
【００５０】
【数２】

なお、需要予測の手法は何でもよく、ニューラルネットワークやＧＭＤＨ（Group Method of Data Handling）、その他の統計的手法等、何でもよい。
【００５１】
次に、最適運用計画演算手段２５による最適運用計画を作成する演算手法を説明する。この演算手法は、以下説明する最適化問題を解くことによって得られる。
【００５２】
ここで、最適運用計画演算手段２５は、ヒューマンインターフェース部２２から入力された項目、すなわち、運用に要するエネルギー最小化やピークカット、機器の連続運転といった上水道プラントの運用上、考慮されるべき項目を加味し、最適な運用計画を作成することを目的としている。この考慮されるべき項目の個数に制限はないが、例えば、以下の３つをあげてみる。
【００５３】
（１）供給（処理）流量変更の最小化
（２）消費電力量のピークカット
（３）配水池目標水位誤差の最小化
これらの項目を満足するため、オペレータが入力した優先度を反映した運用計画最適化問題を次のように定式化する。
【００５４】
［目的関数］

［制約条件］
需要予測値＝配水池水位変化×配水池断面積+浄水場からの送水流量 （４）
配水池下限値＜配水池水位＜配水池上限値 （５）
なお、上記説明は浄水場２から配水池４，１０へ浄水を供給する場合の送水最適運用計画（供給最適運用計画）について行ったが、浄水場内最適運用計画（処理施設内最適運用計画）や取水最適運用計画（取り入れ量最適運用計画）についても、同様の定式化を行うことにより解くことができる。ただし、浄水場内最適運用計画の場合は、需要予測値に代って、作成された送水最適運用計画を用い、取水最適運用計画では、需要予測値に代って、作成された浄水場内最適運用計画を用いる。
【００５５】
以下の説明は、送水最適計画について記述するが、浄水場内最適運用計画や取水最適運用計画についても同様に適用できる。また、プラント全体の運用計画最適化問題は、需要量や配水池などの容量、供給量に相当する送水量等を、後述するように、それぞれベクトル（マトリックス）としてまとめて考え、同様に前述の最適化方法を適用することによって定義することができる。そして、これに対して最適化演算を行なえば、上水道プラント全体の最適運用計画を作成することができる。その際、それぞれの制約条件を考慮して、最適運用計画評価手段３２によってプラント全体を評価する。
【００５６】
ここで、図２で示した送水ポンプ３が可変速ポンプであったり、流量調整バルブ９の開度調整によって送水する場合、定式化された最適化問題における目的関数や制約条件は、離散値ではなく連続値で与えることができる。これまでは、流量の変動を離散的に表現して最適化演算を行なってきたが、最適運用計画演算手段２５における連続値最適計画演算手段２７では、連続的に変化する流量を扱うことのできる実数値遺伝的アルゴリズム（ＲＧＡ：Real Genetic Algorithm） によって上述の最適化問題を解く。
【００５７】
以下に、そのアルゴリズムを示す。
【００５８】
＜ＲＧＡアルゴリズム＞
（STEP-１）初期個体群の生成
ランダムにｎ個の実数型コーディングによる個体を発生させる。
【００５９】
（STEP-２）各個体の評価
制約条件を満足する個体がｎ個生成できたら、各個体の適応度及びその世代での平均適応度を計算する。
【００６０】
（STEP-３）淘汰処理
個体群中の最大適応度に対して、ある比率より大きい適応度を持つ個体を淘汰する。制約条件を満足しない個体が存在した場合も淘汰する。
【００６１】
（STEP-４）増殖処理
淘汰された個体の数だけ、適応度最小の個体を増殖させる。
【００６２】
（ＳＴＥＰ−５）交叉処理
ランダムにペアリングを行う。ペアリングは全個体数に対する割合（交叉確率）分だけ行う。交叉オペレータは様々だが、ここではＢＬＸ−α法を採用する。
【００６３】
（STEP-６）終了判定処理
上記（STEP-２）〜（STEP-５）を繰り返す。ただし、（STEP-２）では、その世代における平均適応度が前２世代の平均適応度と比較して、ある任意に設定する値ε以下又はある繰り返し回数以上であればアルゴリズムを終了する。
【００６４】
以下、上述したRGAアルゴリズムを、図２のプラントに適用した場合を説明する。
【００６５】
浄水場２から配水池４，１０への送水流量の変動を、１／０の遺伝子列ではなく、実数の変数ベクトル［ｘ_１，ｘ_２，・・・，ｘ_Ｎ］を使って表現し、２４時間の計画の場合は次のように１個体を定義する。
【００６６】
＜個体＞
【表１】

すなわち、ｘ_１は１時における送水流量［ｍ^３／ｈ］を意味する。このような変数ベクトルを個体と呼び、適当な数の個体群を設定する。
【００６７】
なお、（STEP-１）における初期個体群の生成において、初期個体をランダムに生成すると効率がよくないため、次のような処理を施すとよい。
【００６８】
すなわち、変数ベクトルの要素をすべて等しくしたものを生成可能であるとする。つまり、Ｎ時まで一定の送水流量である流量計画を初期個体として生成させる。その際に、制約条件を満足するものが一つもなかった場合でも、制約条件を無視し、制約違反の個体に罰則（重み）を付加させることによって、それぞれの個体を評価し、より優秀な個体を初期個体として生成させる。
【００６９】
なお、初期個体は、過去有効であった運用計画から抽出することで与えてもよい。さらに、初期個体を、ヒューマンインターフェース２２から入力可能としてもよい。
【００７０】
適応度（ある世代で、どの程度問題に対して適応しているかの指標）としては、式（３）で示した目的関数の逆数を考えることとする。ただし、適応度はこれに限定されるものではない。
【００７１】
解の探索性能がよいＲＧＡを実現するためには交叉オペレータの設計が重要である。ＲＧＡに対しては数多くの交叉オペレータが提案されているが、ここでは単峰性・多峰性関数の両方に対して良好な探索性能を示すＢＬＸ−αと呼ばれる交叉オペレータを説明する。図３はこの概念図である。
【００７２】
図３において、親個体をｐ_１，ｐ_２とするとき、子個体ｃを次式にしたがって生成する。
【００７３】
ｃ_ｉ＝ｕ｛min（ｐ_１ｉ，ｐ_２ｉ）−αＩ，max（ｐ_１ｉ，ｐ_２ｉ）＋αＩ｝（６）
Ｉ＝｜ｐ_１ｉ−ｐ_２ｉ｜
上式において、ｃ_ｉ，ｐ_１ｉ，ｐ_２ｉは、それぞれｃ，ｐ_１，ｐ_２の要素であり、ｕ（ｘ，ｙ）は区間［ｘ，ｙ］の一様乱数とする。つまり、子個体は座標軸に平行な辺からなる長方形の内部に一様分布する。また、αは大域的探索能力と局所的探索能力との重み付けの役割を果たす。つまり、図３に示すように、α＝０の場合は、子が親同士の間にできるため、収束は早いが探索範囲は狭くなる。
【００７４】
なお、従来の０−１遺伝子のＧＡでは、突然変異処理を施すのが一般的であるが、今回採用したＢＬＸ−αの交叉オペレータには、ｕ（ｘ，ｙ）によってその役割を果たすことができるので、突然変位処理は行なわない。
【００７５】
次に、目的関数や制約条件の実装例を示す。なお、目的関数や制約条件は、これに一意に定まるものではない。
【００７６】
目的関数：Ｊ＝Ｊ_ｓ＋Ｊ_ｐ＋Ｊ_ｈ
（１）流量変化の最小化 Ｊ_ｓ
・バルブ９の開度調整による流量制御の場合
【数３】

・可変速ポンプ３による流量制御の場合
【数４】

ｎ（ｘ_ｋ）：計画流量ｘ_ｋの時のポンプ運転台数
ここで、ｎ（ｘ_ｋ）の決定方法は様々だが、ここでは以下の方法を例として記す。
【００７７】
図４は、可変速ポンプのＱ−Ｈ（吐出量−全揚程）曲線において、２台のポンプが並列運転可能な場合を示している。同図において、計画流量が、１台の効率１００％のときの流量Ｑ_Ｈ１までは１台運転を基準とし、それ以上のＱ_Ｈ２までの計画流量の場合は２台並列運転に切り換える。つまり、次式のように表される。
【００７８】
【数５】

ここで、ｍはｍ台並列運転であることを表し（図４の場合はｍ＝２）、Ｑ_Ｈｉはｉ台並列運転時の効率が１００％のときの吐出量を表す。
【００７９】
（２）消費電力の最小化 Ｊ_ｐ
【数６】

（３）配水池４，１０目標水位誤差の最小化 Ｊ_ｈ
【数７】

なお、ｗｉ（ｋ）（ｉ＝１，２，３，４）は、それぞれの項目に対する時刻ｋの重み係数である。
【００８０】
制約条件：
【外１】

配水池の水位変化を同時に考慮すると、これらの関係は次のようになる。
【００８１】
【数８】

ただし、Ａは配水池の底面積［ｍ^２］を表す。
【００８２】
・各時刻における各配水池の水位
ｈ_ＬＷＬ≦ｈ_ｋ≦ｈ_ＨＷＬ （１２）
ｈ_ＬＷＬ，ｈ_ＨＷＬは、それぞれ配水池の上下限水位を表す。
【００８３】
・一度に変更できるポンプ運転台数
｜ｎ（ｘ_ｋ）−ｎ（ｘ_ｋ−１）｜≦Ｒ （１３）
Ｒは一度に変更できるポンプ運転台数とする。
【００８４】
・バルブの静水圧
静水圧力Ｐが、その温度における飽和蒸気圧Ｐｖまで低下すると、その部分で水は沸騰してしまい、キャビテーション現象を引き起こしてしまう。キャビテーション現象は、液体を扱う機器においては不可避の問題であり、制約条件は以下のようになる。
【００８５】
Ｐ＞Ｐｖ （１４）
このように、最適化問題を定式化し、これを解くことにより最適な運用計画を作成することができる。
【００８６】
また、上水道プラント全体を対象とする運用計画を得る場合は、変数ベクトルを変数マトリックスとして定義することにより、導出可能となる。つまり、浄水池１４から送水する配水池がｍ池ある場合、以下のようなマトリックスを導入する。
【００８７】
【表２】

この場合、制約条件は、変数マトリックスに与えられるのではなく、マトリックスの行（対象）毎に与えられることになる。
【００８８】
なお、目的関数や制約条件が離散値で定義される場合にも、上述したＲＧＡを適用することができる。この場合、目的関数のＪ_ｓの項は、（８）式の２項目のみを採用する。ただし、ｎ（ｘ_ｋ）は（９）式に定まるものではない。
【００８９】
なお、図１で示した離散値最適計画演算手段２８では、組合わせ最適化問題として定式化し、代表的な解法を用いることもできる。この場合、最適化問題を解く解法は遺伝的アルゴリズムに限定されず、数理計画法（列挙法の１種である分岐限定法や動的計画法等）、タブサーチ法、シミュレーテッドアニーリング法、免疫的アルゴリズム法等、何でもよい。
【００９０】
また、最適運用計画演算手段２５は、送水流量等の計画変数を連続量として扱うか離散量として扱うか、切換手段２６を設けたことにより、対象によって選択切換することができ、あらゆる対象に対し運用計画を作成することができる。すなわち、計画変数が連続量の制御対象の場合は、切換手段２６により予測された需要量を連続値最適計画演算手段２７に与え、計画変数が離散値の制御対象の場合は、予測された需要量を離散値最適計画演算手段２８に与えるようにしている。
【００９１】
また、計画変数が連続量及び離散量の双方を含む場合は、予測された需要量を、連続値最適計画演算手段２７と離散値最適計画演算手段２８の双方に与えるようにしてもよい。この場合、切換手段２６を省略し、需要予測手段２４の出力を、連続値最適計画演算手段２７と離散値最適計画演算手段２８の双方に直結してもよく、あるいは、切換手段２６を、図示のいずれかの端子に選択接続する機能の他に、双方の端子に接続する、所謂双投形の切換機能を持つように構成してもよい。
【００９２】
最適運用計画演算手段２５の運用計画評価手段３２は、前述した適応度をそれぞれの対象毎に評価するものであり、図示のように、連続値最適計画演算手段２７や離散値最適計画演算手段２８とのループを形成している。そして、評価された計画が前述した終了判定を満たすなら、その運用計画をプロセス入出力部２１を介してプラントに出力させる。
【００９３】
ここで、送水計画評価手段２９では送水計画に関する適応度を評価し、浄水場内運用計画評価手段３０では浄水場２内の運用に関する適応度を評価し、取水運用計画評価手段３１では各取水源７，８から最適に原水を取水する計画を評価する。また、運用計画評価手段３２は、前述したように、送水計画、浄水場運用計画、取水運用計画を一括して評価することもできる。
【００９４】
さらに、運用計画評価手段３２は、図２の送水ポンプ３が固定速ポンプと可変速ポンプとから構成されている場合等、連続量と離散量とが複合している場合は、連続値最適計画演算手段２７と離散値最適計画演算手段２８とからの出力を総合して評価する。
【００９５】
このように、遺伝子として、何時にどれだけの供給量が必要なのかを実数値の変数ベクトルとして表現し、実数値遺伝的アルゴリズムを適用することによって、離散的な定義では表現できなかった運用計画が考慮でき、最適な供給量を導出することができる。
【００９６】
なお、上記実施の形態では、対象プラントとして上水道プラントを例示したが、これに限定されるものではなく、例えば、発電所、ガス、及び熱（冷水、温水）などの供給プラント等にも適用することができる。
【００９７】
【発明の効果】
本発明によれば、プラントの運用計画を需要予測に基づいて最適化するに当り、計画変数が離散量ではなく連続量として与えられるプラントにおいても、連続な実数値の計画を考慮することで、バルブの開度調整や、可変速ポンプによる連続的に変化する流量等を、離散的に定義することなく扱うことができる最適運用計画の意思決定を行ったり、支援だけでなくプラントを自動調整することができる。
【図面の簡単な説明】
【図１】本発明によるプラント運用最適化制御装置の一実施の形態を示すブロック図である。
【図２】本発明の対象となるプラント例を示す概念図である。
【図３】上記一実施の形態に用いる交叉オペレータＢＬＸ−α法の概念図である。
【図４】同上一実施の形態に用いる可変速ポンプのＱ−Ｈ特性を説明する特性図である。
【符号の説明】
２ 処理施設の一例である浄水場
４，１０ 分配施設の一例である配水池
７，８ 供給源の一例である取水源（河川、地下水）
２０ プラント運用最適化制御装置
２５ 最適運用計画演算手段
２６ 切換手段
２７ 連続値最適計画演算手段
２８ 離散値最適計画演算手段
３２ 運用計画評価手段[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a plant operation optimization control apparatus that creates an optimum planned value for a required value in various plants.
[0002]
[Prior art]
In various plants, optimal control is performed to create an optimal planned value for a required value. For example, there is a water supply plant that produces or supplies purified water for the water distribution demand of the water supply, or takes both of them. In such a water supply plant, it is necessary to create an operation plan for supplying clean water safely and stably by efficiently using the capacity of the reservoir for the demand for water supply that changes every moment. For this reason, plant operation optimization control that automatically creates a plant operation plan such as a water transfer plan from a water treatment plant to a distribution pond, a water purification production plan at the water purification plant, a raw water intake plan, and a plan for the entire plant including those plans A device is required.
[0003]
Conventionally, in such a plant operation optimization control apparatus, plan variables such as a purified water supply amount and a treated water treatment amount are defined as discrete values, and a mathematical programming method or a combination optimization method (for example, a genetic by a 0-1 gene) is defined. An optimal plan is calculated using an algorithm, and an operation plan for a water supply plant is created (for example, see Patent Document 1).
[0004]
As described above, even when a genetic algorithm based on the 0-1 gene, which is one of the combination optimization methods, is used, the individual is usually expressed by a discrete variable. In this case, the plan such as the supply amount and the treatment amount must be dealt with discretely, and the continuous water amount change due to the valve and the variable speed pump cannot be considered.
[0005]
[Patent Document 1]
Japanese Patent Laid-Open No. 8-134962
[0006]
[Problems to be solved by the invention]
In this way, even in a target that can take into account continuous flow fluctuations due to valves and variable speed pumps, the fluctuations are discretely defined, so unnecessary unit control occurs during control, and unnecessary power consumption There was a problem that was consumed.
[0007]
In order to solve this problem, it is necessary to create a plant operation plan by treating the plan variable as a continuous value, not as a discrete value, based on a demand forecast for predicting water distribution demand. Realizing this is an important issue for improving the efficiency of water operations, such as suppressing unnecessary number control.
[0008]
  The object of the present invention is to take into account continuous real-valued planning, even in plants where the planning variables are defined as continuous rather than discrete.ReamAn object of the present invention is to provide an optimum plant operation control apparatus capable of deriving an optimum operation plan that can handle a continuously changing amount as a plan variable.
[0009]
[Means for Solving the Problems]
  The plant operation optimization control device according to the present invention performs a predetermined process at a processing facility on an object taken from a supply source, and supplies the target object to the predetermined supply target facility from the processing facility. A demand control means for calculating a demand forecast value for the next N hours from plant data and past demand record data input from the plant side, and future N from the processing facility to the supply target facility. When supply plan variables for creating a supply plan for time are given as continuous quantities, a supply optimal operation plan that satisfies the conditions required for the demand forecast value is formulated with objective functions and constraints. As an optimization problem, these objective functions and constraints are given as continuous values, and individuals with n real type coding are generated. , The optimal operation plan arithmetic means having a continuous value optimal design computing means for creating the objective function so as to minimizeAnd a variable speed pump for supplying the target object from the processing facility to a predetermined supply target facility.,The discharge variable of the variable speed pump given as a continuous variable is a design variable. The n individuals are variable vectors of real numbers of the plan variable for N hours. Evaluate the number of operating variable speed pumps by considering the QH characteristics of the variable speed pumpsIt is characterized by that.
[0010]
  Further, the plant operation optimization control apparatus according to the present invention has a flow rate control valve provided between the processing facility and a predetermined supply target facility.,The opening of the valve is given as a continuous variable as a planned variable, the n individuals are variable vectors of the real number of the planned variable for N hours, and the absolute constraint condition is the static water pressure of the valve> saturated vapor pressure Suppress cavitation of the valveCharacterized by
[0011]
  In the present invention, the continuous value optimum plan calculation means in the optimum operation plan calculation means is based on a real value genetic algorithm based on the prepared supply optimum operation plan instead of the demand forecast value.An optimal operation plan within the processing facility may be created.
[0012]
  In the present invention,The continuous value optimal plan calculation means in the optimal operation plan calculation means equalize and quantitate the amount taken from the supply source by the real-valued genetic algorithm based on the optimized in-process operation plan created instead of the demand forecast value. It may be possible to create an intake amount optimum operation plan that is optimized so as to be optimized.
[0013]
  In the present invention,The optimal operation plan calculation means includes a discrete value optimal plan calculation means for handling a discrete quantity and solving an optimization problem when the target includes a plan variable given as a discrete quantity. It is good also as a structure which provided the operation plan evaluation means which monitors the output with a calculating means and evaluates both these outputs collectively.
[0014]
  In the present invention, the optimum operation plan calculation means is:It is equipped with discrete value optimal plan calculation means that treats the planning variables given as discrete quantities as discrete quantities and solves optimization problems, and switching means that switches whether the plan variables are treated as continuous quantities or discrete quantities depending on the target It is good also as a structure.
[0015]
  In the present invention,In the case where there are a plurality of supply target facilities, there is provided an evaluation means for comprehensively evaluating the supply amounts to the respective supply target facilities with the objective function of smoothing the supply amounts from the processing facilities to the respective supply target facilities. Good.
[0016]
  In the present invention,The initial individual of the continuous value optimum plan calculation means may be extracted from the operation plan that has been valid in the past.
[0017]
  Further, in the present invention, the initial individual of the continuous value optimal plan calculation means isIt may be possible to input from a human interface.
[0018]
  Furthermore, the present invention provides a demand prediction means for calculating a demand forecast value for N hours from the plant data inputted from the plant side and past demand record data, and the whole plant from the supply source to the supply target facility. When planning variables of each part constituting the plant for creating an operation plan for N hours in the future are given as continuous quantities, each variable vector by real numbers for each planning variable is defined as a variable matrix, and the demand forecast value Optimal operation to create an operation plan for the entire plant that satisfies the required conditions as an optimization problem formulated with an objective function and constraints, and to create the objective function to be minimized by a real-valued genetic algorithm A plan calculation means, and the constraint condition may be given to each target of each part constituting the variable matrix.
[0020]
In these inventions, when a plan variable is given as a continuous quantity, an optimal operation plan that satisfies the required items is an optimization problem formulated with an objective function and constraint conditions, and these objective functions and constraint conditions are continuously generated. It is given by a value, n individuals are generated by real number type coding, and created by a real-valued genetic algorithm so as to minimize the objective function.
[0021]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, an embodiment of a plant operation optimization control apparatus according to the present invention will be described with reference to the drawings.
[0022]
First, referring to FIG. 2, the configuration of a waterworks plant will be described as an example of a plant that is the subject of the present invention.
[0023]
In FIG. 2, raw water (target object) is taken into the water purification plant (treatment facility) 2 by the intake pumps 1 and 15 from the river 7 and the groundwater source 8 serving as the intake source (supply source). The water purification plant 2 is provided with a flocculation / sedimentation / filtration facility 13 as a treatment facility and a water purification pond 14 for temporarily storing the filtered water.
[0024]
The purified water that has been subjected to the water purification treatment at the water purification plant 2 is supplied to a distribution reservoir (supply target facility) 4 by a water pump 3. Alternatively, the other distribution reservoirs (supply target facilities) 10 are supplied by a natural flow-down method using gravity. In this case, a valve 9 for adjusting the flow rate is provided in the supply path. From these distribution reservoirs 4, 10, water is distributed to corresponding distribution districts 6, 12 by distribution pumps 5, 11.
[0025]
In the water purification plant 2, the raw water is subjected to flocculation treatment, chemical precipitation, and further filtration treatment to produce purified water. In this case, in order to stabilize the quality of the purified water, to make the amount of chemicals such as flocculant and hypochlorous acid used in the purified water production appropriate, and to carry out stable purified water production, the amount of treated water is as constant as possible. It is necessary to make quantity. For this purpose, it is required to smoothen the supply amount from the clean water reservoir 14 to the distribution reservoirs 4 and 10 by efficiently utilizing the capacity of the distribution reservoirs 4 and 10. Therefore, it is necessary to optimize the operation plan of the water pump 3 and the flow rate adjusting valve 9.
[0026]
In response to such demands, conventionally, the planned flow rate by the water pump 3 and the flow rate adjusting valve 9 is discretely defined, and the water plan is optimized using a genetic algorithm based on the 0-1 gene.
[0027]
However, when the water supply pump 3 is a variable speed pump, it is effective not to define the water supply flow rate discretely but to consider it as a continuous amount from the viewpoint of smoothing the water supply amount. In addition, water is supplied to the distribution reservoir 10 under a natural flow using the position lower than the clean water reservoir 14, but in this case also, the plan is optimized by setting the opening of the flow rate adjusting valve 9 as a continuous value. It is necessary to make it.
[0028]
On the other hand, in order to stably perform water purification treatment at the water purification plant 2, it is necessary to plan the operation of the filter basin efficiently in consideration of the timing of the filter basin washing and the residence time based on the obtained water transmission plan. There is. In addition, when the raw water to be purified is taken from the appropriate water sources 7 and 8, it is necessary to take as much water as possible from the viewpoint of water rights.
[0029]
Therefore, the operation at the water purification plant 2, which is a flocculation / chemical precipitation / filtration facility, and the operation of the intake pumps 1 and 15 for taking raw water from the water source need to be optimally planned as well.
[0030]
In addition, the waterworks plant of FIG. 2 is an example, and the actual waterworks plant includes the number of water purification plants 2 and distribution reservoirs 4 and 10, the number and types of water intake sources, and the number and types of water supply pumps 3 and water distribution pumps 5 and 11. Is also very colorful.
[0031]
  FIG. 1 shows a plant operation optimization control device 20 that effectively acts on the above-mentioned demand. In FIG. 1, reference numeral 21 denotes a process input / output unit, which is provided between the plant operation optimization control device 20 and the plant shown in FIG. 2 and inputs various measurement signals and state signals from the plant and optimizes the plant operation. The control output from the control apparatus 20 is output to the plant side.22Is a human interface unit, which is provided between the plant operation optimization control device 20 and the operator, and outputs display information from the plant operation optimization control device 20 or operates to the plant operation optimization control device 20. Communicate input information by employees.
[0032]
Next, each function realization means in the plant operation optimization control apparatus 20 will be described. Reference numeral 23 denotes plant data storage means for storing all or part of various types of plant data input from the process input / output unit 21. Reference numeral 24 denotes a demand prediction unit that predicts the demand for purified water in the water distribution zones 6 and 12 using the plant data stored in the plant data storage unit 23.
[0033]
Reference numeral 25 denotes an optimum operation plan calculation means that creates an optimum operation plan for smoothing the supply amount and the treated amount of purified water based on the demand prediction value obtained by the demand prediction means 24. The human interface 22 actually exchanges information with the optimum operation plan calculation means 25. As described above, the human interface 22 displays necessary data and sends it to each means such as the demand prediction means 24. The operation / control signal can be issued.
[0034]
Reference numeral 27 denotes a continuous value optimum plan calculation means which can handle and calculate a flow rate that changes continuously by considering a continuous real value plan. That is, when the supply plan variable is given as a continuous quantity like a flow rate, a supply optimum operation plan that satisfies the required conditions (for example, smoothing the supply quantity) is created for this demand forecast value. That is, as will be described later, an optimization problem formulated with an objective function and constraint conditions is given, and these objective functions and constraint conditions are given as continuous values to generate individuals by n real-type coding, and real-valued genetics An algorithm is used to minimize the objective function.
[0035]
Reference numeral 28 denotes a discrete value optimum plan calculation means, which can be calculated by handling discretely defined flow rates in consideration of a discrete value operation plan. That is, when a plan variable given as a discrete quantity is included as in the case of a unit control pump, the discrete quantity is handled by 0-1 genes and the optimization problem is solved by a genetic algorithm as in the conventional case.
[0036]
26 is a calculation switching means, the common input terminal is connected to the demand prediction means 24, one output terminal is connected to the continuous value optimum plan calculation means 27, and the other output terminal is connected to the discrete value optimum plan calculation means 28, respectively. It is connected and switches whether the plan variable is handled as a continuous quantity or a discrete quantity depending on the object.
[0037]
An operation plan evaluation unit 32 includes a water transmission plan evaluation unit 29, a water treatment plant operation plan evaluation unit 30, and a water intake operation plan evaluation unit 31. The operation plan evaluation means 32 inputs the calculation results from the continuous value optimum plan calculation means 27 and the discrete value optimum plan calculation means 28 via the evaluation switching means 33 and 34 and forms a loop, which plan is better? Can be evaluated and selected.
[0038]
The evaluation switching means 33 and 34 are used to switch the range to be planned by the continuous value optimum plan calculation means 27 or the discrete value optimum plan calculation means 28. That is, the following cases (1) and (2) are switched.
[0039]
(1) When planning the water supply plan and water intake plan after considering only the water supply plan, assuming that the water supply plan is a demand forecast.
[0040]
(2) When the operation plan for water transmission, water treatment plant, and water intake is derived all at once.
[0041]
All of the switching of the evaluation switching means 33 and 34 operate independently. After switching, when the input is made to the water supply plan evaluation means 29 (when the right contact point of the switching means 33 is turned on or the left contact point of the switching means 34 is turned on) When it is input to the operation plan evaluation means 32 (when the left contact point of the switching means 33 is turned on or the right contact point of the switching means 34 is turned on), it means (2) described above.
[0042]
The water supply plan evaluation means 29 evaluates the water flow rate plan from the clean water reservoir 14 to the distribution reservoirs 4, 10 created by either of the optimum plan calculation means 27, 28 based on the demand prediction result. In addition, the water treatment plant operation plan evaluation means 30 is a filtration plan in the water treatment plant 2 created by either of the optimum plan calculation means 27 and 28 based on the operation plan evaluated by the water transmission plan evaluation means 29, etc. Evaluate water treatment plans. Further, the intake operation plan evaluation means 31 is based on the operation plan evaluated by the operation plan evaluation means 30 in the water purification plant, and is generated from either intake source 7 or 8 created by either of the optimum plan calculation means 27 and 28. Evaluate the optimal water intake plan.
[0043]
The operation plan evaluated and created by these is output from the process input / output unit 21 to the plant side.
[0044]
In the above configuration, the plant data storage means 23 includes the amount of water purification in the distribution districts 6 and 12 shown in FIG. 2, the operation results of the distribution pumps 5 and 11, the water level of the distribution reservoirs 4 and 10, Data related to operation and control of the water supply plant, such as operation results, operation results in the water purification plant 2, operation results of the intake pumps 1 and 15, and the like are organized for each date, day of the week, etc., and stored together with weather information and the like.
[0045]
Based on the plant data stored in the plant data storage unit 23, the demand prediction unit 24 calculates a demand prediction value for N hours in the future. Based on the obtained demand forecast value, the optimum operation plan calculation means 25 creates an operation plan for the water supply plant. Then, based on the created operation plan, an operation control command for the distribution reservoirs 4, 10 and the water purification plant 2 is transmitted from the process input / output unit 21 to each device of the waterworks plant and controlled.
[0046]
Hereinafter, this demand forecast and water supply plant operation plan creation will be described in detail.
[0047]
  The demand prediction by the demand prediction means 24 may be any method. An example will be described below. 1 hour in the plant data storage means 23everyIt is assumed that the demand result data for one day is filed by day of the week, for example, by holiday, weekday, or holiday. Now, when the day of the week is input from the human interface 22 to the demand prediction unit 24, the average value pattern for one day of the day W is obtained from the demand record in the demand prediction unit. In addition, since the actual demand up to the previous day has been obtained, for example, the following autoregressive model is used to predict the demand for k days.
[0048]
[Expression 1]

A₁, A₂... Are autoregressive parameters, which can be given in advance, or can be subjected to successive least squares estimation (Kalman filter) in real time.
[0049]
Thus, since the predicted value of the day is obtained by the above equation (2) and the pattern for one day of the day of the week W is obtained, when integrating by the ratio of each time zone, 24 hours per hour of the day The predicted value of minutes is obtained as follows.
[0050]
[Expression 2]

The demand forecasting method may be any method, such as a neural network, GMDH (Group Method of Data Handling), or other statistical methods.
[0051]
Next, a calculation method for creating an optimum operation plan by the optimum operation plan calculation means 25 will be described. This calculation method is obtained by solving the optimization problem described below.
[0052]
Here, the optimum operation plan calculation means 25 is an item input from the human interface unit 22, that is, an item to be considered in the operation of the water supply plant such as energy minimization, peak cut, and continuous operation of the equipment. The purpose is to create an optimal operation plan. There is no limit to the number of items to be considered, but the following three are given as examples.
[0053]
(1) Minimization of supply (processing) flow rate change
(2) Peak cut in power consumption
(3) Minimization of reservoir water level error
In order to satisfy these items, the operation plan optimization problem reflecting the priority inputted by the operator is formulated as follows.
[0054]
[Objective function]

[Restrictions]
Demand forecast value = distribution reservoir water level change x distribution reservoir cross-sectional area + water supply flow rate from water treatment plant (4)
Reservoir lower limit <Reservoir water level <Reservoir upper limit (5)
In addition, although the said description was performed about the water supply optimal operation plan (supply optimal operation plan) in the case of supplying purified water from the water purification plant 2 to the reservoirs 4 and 10, the optimal operation plan in a water purification plant (optimum operation plan in a processing facility) The water intake optimum operation plan (intake amount optimum operation plan) can also be solved by performing the same formulation. However, in the case of the optimum operation plan in the water treatment plant, the prepared water supply optimum operation plan is used instead of the demand forecast value, and in the optimum water intake operation plan, the created optimum operation in the water treatment plant is used instead of the demand forecast value. Use a plan.
[0055]
The following explanation describes the optimal water supply plan, but the same applies to the optimal operation plan within the water treatment plant and the optimal intake operation plan. In addition, the operation plan optimization problem for the entire plant is considered as a vector (matrix), as described later, in terms of the volume of demand, the capacity of reservoirs, the amount of water supply corresponding to the supply volume, and so on. It can be defined by applying optimization methods. And if an optimization calculation is performed with respect to this, the optimal operation plan of the whole waterworks plant can be created. At that time, the entire plant is evaluated by the optimum operation plan evaluation means 32 in consideration of each constraint condition.
[0056]
Here, when the water pump 3 shown in FIG. 2 is a variable speed pump or water is supplied by adjusting the opening of the flow rate adjusting valve 9, the objective function and the constraint conditions in the formulated optimization problem are not discrete values. Can be given as continuous values. Until now, optimization calculations have been performed by discretely expressing flow rate fluctuations, but the continuous value optimum plan calculation means 27 in the optimum operation plan calculation means 25 can handle continuously changing flow rates. The above optimization problem is solved by a real genetic algorithm (RGA).
[0057]
The algorithm is shown below.
[0058]
<RGA algorithm>
(STEP-1) Generation of initial population
R individuals are randomly generated by real type coding.
[0059]
(STEP-2) Evaluation of each individual
When n individuals satisfying the constraint conditions are generated, the fitness of each individual and the average fitness of the generation are calculated.
[0060]
(STEP-3) Wrinkle treatment
An individual having fitness greater than a certain ratio with respect to the maximum fitness in the population is selected. Even if there are individuals that do not satisfy the constraints, hesitate.
[0061]
(STEP-4) Proliferation treatment
The number of individuals with the minimum fitness is increased by the number of individuals that have been culled.
[0062]
  (STEP-5) Crossover process
  Pair at random. Pairing is a percentage of the total number of individualsrate) Do it for minutes. Although there are various crossover operators, the BLX-α method is adopted here.
[0063]
(STEP-6) End determination process
Repeat (STEP-2) to (STEP-5) above. However, in (STEP-2), the algorithm is terminated if the average fitness level in that generation is less than a certain arbitrarily set value ε or more than a certain number of iterations compared to the average fitness level of the previous two generations.
[0064]
Hereinafter, the case where the RGA algorithm mentioned above is applied to the plant of FIG. 2 is demonstrated.
[0065]
The fluctuation of the water flow rate from the water treatment plant 2 to the distribution reservoirs 4 and 10 is not a 1/0 gene string but a real variable vector [x₁, X₂, ..., x_N], And in the case of a 24-hour plan, one individual is defined as follows.
[0066]
<Individual>
[Table 1]

That is, x₁Is the water flow rate at 1 o'clock [m³/ H]. Such a variable vector is called an individual, and an appropriate number of individuals are set.
[0067]
In addition, in the generation of the initial population in (STEP-1), since it is not efficient to generate the initial individuals at random, the following processing may be performed.
[0068]
In other words, it is possible to generate a variable vector with all elements equal. That is, a flow rate plan having a constant water supply flow rate until N o'clock is generated as an initial individual. At that time, even if there is no one that satisfies the constraint condition, each individual is evaluated by ignoring the constraint condition and adding a penalty (weight) to the individual in violation of the constraint. Is generated as an initial individual.
[0069]
The initial individual may be given by being extracted from an operation plan that has been valid in the past. Further, the initial individual may be input from the human interface 22.
[0070]
As the fitness (an index of how much the problem is adapted to a problem in a certain generation), the reciprocal of the objective function shown in Equation (3) is considered. However, the fitness is not limited to this.
[0071]
In order to realize RGA with good solution search performance, the design of the crossover operator is important. A number of crossover operators have been proposed for the RGA. Here, a crossover operator called BLX-α that exhibits good search performance for both unimodal and multimodal functions will be described. FIG. 3 is a conceptual diagram of this.
[0072]
In FIG.₁, P₂Then, the child individual c is generated according to the following equation.
[0073]
c_i= U {min (p_1i, P_2i) -ΑI, max (p_1i, P_2i) + ΑI} (6)
I = | p_1i-P_2i｜
Where c_i, P_1i, P_2iAre c and p, respectively.₁, P₂U (x, y) is a uniform random number in the interval [x, y]. That is, the child individuals are uniformly distributed inside a rectangle composed of sides parallel to the coordinate axis. Α plays a role of weighting the global search ability and the local search ability. That is, as shown in FIG. 3, when α = 0, the child can be formed between the parents, so the convergence is fast but the search range is narrow.
[0074]
  In the conventional GA of 0-1 gene, sudden changeDifferentIn general, the processing is performed, but the intersection operator of BLX-α adopted this time can fulfill its role by u (x, y), so sudden displacement processing is not performed.
[0075]
Next, implementation examples of objective functions and constraint conditions are shown. Note that the objective function and constraint conditions are not uniquely determined.
[0076]
Objective function: J = J_s+ J_p+ J_h
(1) Minimizing changes in flow rate J_s
・ For flow control by adjusting the opening of the valve 9
[Equation 3]

・ In case of flow control by variable speed pump 3
[Expression 4]

n (x_k): Planned flow rate x_kNumber of pumps operated at
Where n (x_k) Is determined in various ways, but the following method is used as an example.
[0077]
FIG. 4 shows a case where two pumps can be operated in parallel in the QH (discharge amount-total head) curve of the variable speed pump. In the figure, the flow rate Q when the planned flow rate is 100% efficiency._H1Up to 1 unit operation, and Q_H2If the planned flow rate is up to 2 units, switch to parallel operation. That is, it is expressed as the following formula.
[0078]
[Equation 5]

Here, m represents m units in parallel operation (m = 2 in the case of FIG. 4), Q_HiRepresents the discharge amount when the efficiency during the i-unit parallel operation is 100%.
[0079]
(2) Minimization of power consumption J_p
[Formula 6]

(3) Reservoir 4, 10 Minimization of target water level error J_h
[Expression 7]

Note that wi (k) (i = 1, 2, 3, 4) is a weighting coefficient at time k for each item.
[0080]
Restrictions:
[Outside 1]

Considering the change in the water level of the reservoir at the same time, these relationships are as follows.
[0081]
[Equation 8]

However, A is the bottom area of the reservoir [m²].
[0082]
・ Water level in each reservoir at each time
h_LWL≦ h_k≦ h_HWL                                  (12)
h_LWL, H_HWLRepresents the upper and lower water levels of the reservoirs.
[0083]
・ Number of pumps that can be changed at one time
| N (x_k) -N (x_k-1) | ≦ R (13)
R is the number of pumps that can be changed at one time.
[0084]
  ・ Valve hydrostatic pressure
  When the hydrostatic pressure P decreases to the saturated vapor pressure Pv at that temperature, the water boils in that portion.SistersCause the cavitation phenomenon. The cavitation phenomenon is an unavoidable problem in liquid handling equipment, and the constraint conditions are as follows.
[0085]
P> Pv (14)
Thus, an optimization operation plan can be created by formulating an optimization problem and solving it.
[0086]
Moreover, when obtaining an operation plan for the entire waterworks plant, it can be derived by defining a variable vector as a variable matrix. That is, when there are m distribution ponds that supply water from the clean water reservoir 14, the following matrix is introduced.
[0087]
[Table 2]

In this case, the constraint condition is not given to the variable matrix but given to each row (object) of the matrix.
[0088]
Note that the RGA described above can also be applied when the objective function and the constraint conditions are defined as discrete values. In this case, the objective function J_sOnly the two items of the equation (8) are adopted as the term of (8). Where n (x_k) Is not determined by equation (9).
[0089]
1 can be formulated as a combination optimization problem and a typical solution can be used. In this case, the solving method to solve the optimization problem is not limited to the genetic algorithm, but is mathematical programming (branch and bound method or dynamic programming which is one of enumeration methods), tab search method, simulated annealing method, immune method Any method such as an algorithmic method may be used.
[0090]
In addition, the optimum operation plan calculation means 25 can select and switch depending on the target by providing a switching means 26 to handle the plan variable such as the water flow rate as a continuous quantity or as a discrete quantity. Operation plan can be created. That is, when the plan variable is a continuous quantity control target, the demand amount predicted by the switching means 26 is given to the continuous value optimum plan calculation means 27, and when the plan variable is a discrete value control target, the predicted demand is calculated. The quantity is given to the discrete value optimum plan calculation means 28.
[0091]
When the plan variable includes both continuous and discrete quantities, the predicted demand may be given to both the continuous value optimum plan calculation means 27 and the discrete value optimum plan calculation means 28. In this case, the switching means 26 may be omitted, and the output of the demand prediction means 24 may be directly connected to both the continuous value optimal plan calculation means 27 and the discrete value optimal plan calculation means 28. Alternatively, the switching means 26 is illustrated. In addition to the function of selectively connecting to any one of the terminals, a so-called double throw type switching function of connecting to both terminals may be provided.
[0092]
The operation plan evaluation means 32 of the optimum operation plan calculation means 25 evaluates the above-mentioned fitness for each target, and as shown in the figure, the continuous value optimum plan calculation means 27 and the discrete value optimum plan calculation means 28. And form a loop. If the evaluated plan satisfies the end determination described above, the operation plan is output to the plant via the process input / output unit 21.
[0093]
Here, the water supply plan evaluation means 29 evaluates the fitness related to the water supply plan, the water treatment plant operation plan evaluation means 30 evaluates the fitness related to the operation in the water treatment plant 2, and the water intake operation plan evaluation means 31 evaluates each intake source 7. , 8 to evaluate the plan to take the raw water optimally. Moreover, the operation plan evaluation means 32 can also evaluate a water transmission plan, a water purification plant operation plan, and a water intake operation plan collectively as mentioned above.
[0094]
Furthermore, the operation plan evaluation means 32 is a continuous value optimum plan when the continuous quantity and the discrete quantity are combined, such as when the water pump 3 in FIG. 2 is composed of a fixed speed pump and a variable speed pump. The outputs from the calculation means 27 and the discrete value optimum plan calculation means 28 are evaluated in a comprehensive manner.
[0095]
In this way, an operation plan that could not be expressed in a discrete definition by expressing the amount of supply required as a gene as a real-valued variable vector and applying a real-valued genetic algorithm. Therefore, the optimum supply amount can be derived.
[0096]
In the above embodiment, the water supply plant is exemplified as the target plant. However, the present invention is not limited to this, and for example, the present invention is also applied to a power plant, gas, and a supply plant such as heat (cold water, hot water). be able to.
[0097]
【The invention's effect】
According to the present invention, in optimizing the plant operation plan based on the demand forecast, even in a plant where the plan variable is given as a continuous quantity instead of a discrete quantity, by considering a continuous real value plan, Makes decisions on optimal operation plans that can handle valve opening adjustments, continuously changing flow rates with variable speed pumps, etc. without discrete definitions, and automatically adjusts plants as well as support be able to.
[Brief description of the drawings]
FIG. 1 is a block diagram showing an embodiment of a plant operation optimization control apparatus according to the present invention.
FIG. 2 is a conceptual diagram showing an example of a plant that is an object of the present invention.
FIG. 3 is a conceptual diagram of a cross operator BLX-α method used in the embodiment.
FIG. 4 is a characteristic diagram for explaining a QH characteristic of a variable speed pump used in the embodiment.
[Explanation of symbols]
Water treatment plant which is an example of 2 treatment facilities
4,10 Reservoir as an example of distribution facility
7,8 Intake source (river, groundwater) as an example of supply source
20 Plant operation optimization controller
25 Optimal operation plan calculation means
26 switching means
27 Continuous value optimum plan calculation means
28 Discrete value optimum plan calculation means
32 Operation plan evaluation means

Claims (10)

A plant operation optimization control device that performs predetermined processing at a processing facility on an object taken from a supply source, and supplies the target object to the predetermined supply target facility from the processing facility,
Demand forecasting means for calculating a demand forecast value for the next N hours from plant data inputted from the plant side and past demand record data;
When supply plan variables for creating a supply plan for the next N hours from the processing facility to the supply target facility are given as continuous quantities, an optimal supply operation plan that satisfies the conditions required for the demand forecast value And an optimization problem formulated with the objective function and the constraint conditions, the objective function and the constraint condition are given as continuous values, and n individuals are generated by real type coding. And an optimal operation plan calculation means having a continuous value optimal plan calculation means for creating a minimum value ,
A variable speed pump for supplying the object from the processing facility to a predetermined supply target facility ;
The discharge variable of the variable speed pump given as a continuous variable is a design variable. The n individuals are variable vectors of real numbers of the plan variable for N hours. A plant operation optimization control apparatus characterized in that the number of operating variable speed pumps is evaluated by taking into account the QH characteristics of a variable speed pump . A plant operation optimization control device that performs predetermined processing at a processing facility on an object taken from a supply source, and supplies the target object to the predetermined supply target facility from the processing facility,
Demand forecasting means for calculating a demand forecast value for the next N hours from plant data inputted from the plant side and past demand record data;
When supply plan variables for creating a supply plan for the next N hours from the processing facility to the supply target facility are given as continuous quantities, an optimal supply operation plan that satisfies the conditions required for the demand forecast value And an optimization problem formulated with the objective function and the constraint conditions, the objective function and the constraint condition are given as continuous values, and n individuals are generated by real type coding. And an optimum operation plan calculation means having a continuous value optimum plan calculation means for creating a minimum value.
A valve for flow control is provided between the processing facility and a predetermined supply target facility ,
The opening of the valve is given as a continuous variable as a planned variable, the n individuals are variable vectors of the real number of the planned variable for N hours, and the absolute constraint condition is the static water pressure of the valve> saturated vapor pressure A plant operation optimization control apparatus characterized by suppressing cavitation of the valve . The continuous value optimal plan calculation means in the optimum operation plan calculation means is characterized in that instead of the demand forecast value, an optimum operation plan in the processing facility is created by a real-valued genetic algorithm based on the prepared supply optimum operation plan. The plant operation optimization control apparatus according to claim 1 or 2 . The continuous value optimal plan calculation means in the optimal operation plan calculation means equalizes and quantifies the amount taken from the supply source by the real-valued genetic algorithm based on the optimized operation plan in the processing facility instead of the demand forecast value. The plant operation optimization control apparatus according to claim 3 , wherein an intake amount optimum operation plan to be optimized is created. The optimal operation plan calculation means is
When the target includes a design variable given as a discrete quantity, the discrete value optimal plan calculation means for handling the discrete quantity and solving the optimization problem is provided.
Monitoring the output of the continuous value optimal design calculation means and discrete values optimal design computing means, any of claims 1 to 4, characterized in that a management plan evaluating means for evaluating comprehensively these two output The plant operation optimization control device according to any one of the above. The optimal operation plan calculation means is
It has a discrete value optimal plan calculation means that treats design variables given as discrete quantities as discrete quantities and solves optimization problems,
The plant operation optimization control device according to any one of claims 1 to 4 , further comprising switching means for switching whether the plan variable is handled as a continuous quantity or a discrete quantity depending on a target. When there are multiple supply target facilities, an evaluation means for comprehensively evaluating the supply amount to each supply target facility was provided with the objective function of smoothing the supply amount from the processing facility to each supply target facility. The plant operation optimization control apparatus according to any one of claims 1 to 4 , wherein The plant operation optimization control apparatus according to any one of claims 1 to 4 , wherein an initial individual of the continuous value optimum plan calculation means is extracted from an operation plan that has been valid in the past. The plant operation optimization control apparatus according to any one of claims 1 to 4 , wherein an initial individual of the continuous value optimum plan calculation means can be input from a human interface. A plant operation optimization control device that performs predetermined processing at a processing facility on an object taken from a supply source, and supplies the target object from the processing facility to a predetermined supply target facility,
Demand forecasting means for calculating a demand forecast value for the next N hours from plant data inputted from the plant side and past demand record data;
When planning variables of each part constituting the plant for creating an operation plan for the next N hours of the entire plant from the supply source to the supply target facility are given as continuous quantities, each variable vector by a real number for each planning variable Is defined as a variable matrix, and the operation plan of the entire plant that satisfies the conditions required for the demand forecast value is an optimization problem formulated with an objective function and constraint conditions. And an optimal operation plan calculation means for creating an objective function to minimize it,
The plant operation optimization control device according to claim 1, wherein the constraint condition is given for each target of the respective parts constituting the variable matrix .

Images